chore: bump version to 2.1.24

Merge pull request #2890 from amit-shafnir/worktree-nanoclaw-templates
feat(templates): local template loader, ncl --template, and docs
2026-07-03 18:45:07 +08:00 · 2026-07-02 11:57:00 +00:00 · 2026-07-02 14:56:45 +03:00 · 2026-07-02 12:36:33 +03:00 · 2026-06-30 23:05:45 +03:00 · 2026-06-30 23:04:38 +03:00
47 changed files with 2257 additions and 121 deletions
@@ -0,0 +1,97 @@
+---
+name: learn
+description: "Distill a reusable skill from anything — a directory, a URL, pasted notes, or what you just did together — or refine an existing skill with new learnings. Use when the user says '/learn', 'learn this', 'turn this into a skill', 'capture this workflow', 'make a skill from <source>', or 'improve/update the <name> skill'. Produces or updates a .claude/skills/<name>/SKILL.md authored to NanoClaw's skill guidelines. (This CREATES or REFINES a skill from a source; it does not install existing skills from a registry.)"
+---
+
+# Learn — Distill a Skill from Anything
+
+Turn a source — a directory, a URL, pasted notes, or the work just done in this conversation — into a clean, reusable NanoClaw skill. The output is a new `.claude/skills/<name>/SKILL.md` (plus optional `scripts/`, `references/`, `templates/`) authored to the project's skill guidelines.
+
+This skill is **instruction-only**: it uses the tools you already have (`Read`, `Grep`, `Glob`, `WebFetch`, `Write`) — there is no separate distillation engine and no reach-ins into core code.
+
+## When to use
+
+Invoke when the user wants to *capture* a workflow as a reusable skill:
+
+- `/learn <path>` — read a project/dir and build a skill for working with it
+- `/learn <url>` — read docs / an API page and build a usage skill
+- `/learn what we just did` — distill the current conversation's workflow
+- `/learn` + pasted notes — turn notes into a structured skill
+
+If the user instead wants to *find and install* an existing community skill, that is a different task — this skill **creates** new skills, it does not import them.
+
+## Workflow
+
+### 1. Identify the source — and whether this is a new skill or a refine
+- A **path** → read the code/files.
+- A **URL** → fetch and read the page.
+- **"what we just did" / "this"** → use the current conversation as the source.
+- **Pasted text** → use it directly.
+
+Then check `.claude/skills/` for an existing skill that already covers this topic (the user may name it, e.g. *"update the wow-on-steam-deck skill"*, or the subject may obviously match one). **If one exists, this is a REFINE, not a fresh create** — go to step 4's "Refining" branch.
+
+If it is ambiguous what the skill should *do*, ask one clarifying question before proceeding.
+
+### 2. Gather the material
+- **Path:** `Glob` the structure, `Read` the key files, `Grep` for the important entry points. Read enough to understand the *repeatable procedure*, not every line.
+- **URL:** `WebFetch` the page; pull out the concrete commands/steps, not the prose.
+- **Conversation:** re-read what was actually done — the commands, the gotchas, the decisions — and keep the parts that generalize.
+
+### 3. Distill — find the reusable procedure
+Strip the one-off specifics; keep the *repeatable* shape. A good skill answers: *"Next time someone needs to do X, what are the exact steps, files, commands, and gotchas?"* Capture:
+
+- the trigger / when-to-use,
+- the step-by-step procedure (commands, file paths, decision points),
+- the non-obvious **gotchas** that were hit — usually the most valuable part,
+- any scripts or templates worth shipping alongside.
+
+### 4. Author the SKILL.md
+
+**Refining an existing skill?** First `Read` the current `.claude/skills/<name>/SKILL.md`, then *update it in place* — do not blindly overwrite:
+- Keep what is still correct; weave the new learnings into the right sections.
+- **Dedupe** — don't append a near-duplicate step or a second gotcha that says the same thing.
+- Correct anything the new source proves stale (a changed path, command, or flag).
+- Preserve the existing `name`/folder and overall structure; the diff should read as a focused improvement, not a rewrite.
+
+**New skill?** Write `.claude/skills/<kebab-name>/SKILL.md`.
+
+**Frontmatter (required):**
+
+```yaml
+---
+name: <kebab-case, matches the folder>
+description: "<what it does + when to use it + likely trigger phrases>"
+---
+```
+
+`description` is what the agent reads to decide relevance — make it concrete and include the phrases a user would actually say.
+
+**Body:** open with one paragraph on what the skill does, then a `## When to use` section and a `## Workflow` of numbered steps (the actual procedure). Use tables for command/file references, and add a short examples or troubleshooting section when the gotchas warrant it.
+
+**House authoring rules (from `docs/skill-guidelines.md`):**
+
+- **Additive, minimal reach-ins** — prefer adding files; make the *smallest possible* edit to existing code, and only via single-line calls into skill-owned functions.
+- **Instruction-only when possible** — if Claude can do it by following prose plus existing tools, ship no code. These are the easiest skills to maintain and to merge.
+- If apply leaves anything behind, ship a **`REMOVE.md`** that fully reverses every change (no soft-disabled/commented-out removals).
+- If the skill adds an integration point in core code, add a **test that goes red if the wiring is deleted or drifts**.
+- Anti-patterns to avoid: separate `VERIFY.md` files, incomplete cleanup, raw SQL against core DBs, branch merges (use additive fetch), hand-maintained duplicate copies.
+
+### 5. Place and verify
+- Write into `.claude/skills/<name>/`; confirm the folder name matches the `name` frontmatter and the YAML parses.
+- If feasible, dry-run the procedure the skill describes to confirm it is correct.
+- Tell the user the skill exists and how to invoke it (`/<name>`).
+
+## Example
+
+`/learn what we just did` after a multi-step setup:
+
+1. Re-read the conversation's commands and gotchas.
+2. Distill the repeatable procedure.
+3. Write `.claude/skills/<topic>-setup/SKILL.md` with the steps, file paths, and the gotchas hit along the way.
+4. Report: *"Created `/<topic>-setup` — invoke it next time to repeat this."*
+
+## Notes
+
+- Keep skills **focused** — one capability per skill (mirrors the project's "one change per PR" rule).
+- The most valuable content is the **gotchas**, not the happy path.
+- This skill is prose and safe to re-run — use it again to refine an existing skill.
@@ -280,6 +280,7 @@ This project uses pnpm with `minimumReleaseAge: 4320` (3 days) in `pnpm-workspac
 | [docs/customizing.md](docs/customizing.md) | Short intro to customizing via skills |
 | [docs/skills-model.md](docs/skills-model.md) | The skills model in full: recipes, tests, upgrades, migrations |
 | [docs/skill-guidelines.md](docs/skill-guidelines.md) | Authoritative checklist for writing a skill |
+| [docs/templates.md](docs/templates.md) | Agent templates: what they are, stamping via `ncl groups create --template` + the setup wizard, the OneCLI/MCP-credential model, supported providers, and how to contribute one |

 ## Container Build Cache

@@ -125,6 +125,10 @@ Instructions here...
 - Put code in separate files, not inline in the markdown
 - See the [skills standard](https://code.claude.com/docs/en/skills) for all available frontmatter fields

+## Templates
+
+Agent templates (reusable bundles of instructions + MCP servers + skills) ship in the separate [`nanocoai/nanoclaw-templates`](https://github.com/nanocoai/nanoclaw-templates) repo, not this one. Contribute them there via PR (its README has the anatomy and checklist). For how templates load and the OneCLI credential model, see [docs/templates.md](docs/templates.md).
+
 ## Testing

 Test your contribution on a fresh clone before submitting. For skills, run the skill end-to-end and verify it works.
@@ -82,6 +82,7 @@ See [docs/v1-to-v2-changes.md](docs/v1-to-v2-changes.md) for what's different an
 - **Web access** — search and fetch content from the web
 - **Container isolation** — agents are sandboxed in Docker (macOS/Linux/WSL2), with optional [Docker Sandboxes](docs/docker-sandboxes.md) micro-VM isolation or Apple Container as a macOS-native opt-in
 - **Credential security** — agents never hold raw API keys. Outbound requests route through [OneCLI's Agent Vault](https://github.com/onecli/onecli), which injects credentials at request time and enforces per-agent policies and rate limits.
+- **Agent templates**: stamp a ready-to-run agent (instructions + MCP tools + skills, no secrets) from a reusable bundle, via the setup wizard or `ncl groups create --template <ref>`. Load from the [public library](https://github.com/nanocoai/nanoclaw-templates), a local folder, or any git repo. See [docs/templates.md](docs/templates.md).

 ## Usage

@@ -5,8 +5,8 @@
    "": {
      "name": "nanoclaw-agent-runner",
      "dependencies": {
-        "@anthropic-ai/claude-agent-sdk": "^0.3.170",
-        "@anthropic-ai/sdk": "^0.100.0",
+        "@anthropic-ai/claude-agent-sdk": "^0.3.197",
+        "@anthropic-ai/sdk": "^0.108.0",
        "@modelcontextprotocol/sdk": "^1.29.0",
        "cron-parser": "^5.0.0",
        "zod": "^4.0.0",
@@ -19,25 +19,25 @@
    },
  },
  "packages": {
-    "@anthropic-ai/claude-agent-sdk": ["@anthropic-ai/claude-agent-sdk@0.3.170", "", { "optionalDependencies": { "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.170", "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.170", "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.170", "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.170", "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.170", "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.170", "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.170", "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.170" }, "peerDependencies": { "@anthropic-ai/sdk": ">=0.93.0", "@modelcontextprotocol/sdk": "^1.29.0", "zod": "^4.0.0" } }, "sha512-pAvhfk+iTodXZ6RF18Kz7BEUWFjL7EcR3tKuhUNdPpE1NAYCR3mSHGbafi72JsrNwKEDIs7FU31z3fqhwy8QzA=="],
+    "@anthropic-ai/claude-agent-sdk": ["@anthropic-ai/claude-agent-sdk@0.3.197", "", { "optionalDependencies": { "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.197", "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.197", "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.197", "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.197", "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.197", "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.197", "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.197", "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.197" }, "peerDependencies": { "@anthropic-ai/sdk": ">=0.93.0", "@modelcontextprotocol/sdk": "^1.29.0", "zod": "^4.0.0" } }, "sha512-XNIi8W1tb+QfMkcK+5kepOC6BsxG8wtupd72H+pIPzIJypVQhHy7FoX+KBMtTRYwtl+5dsjKyABhjWXebeUilw=="],

-    "@anthropic-ai/claude-agent-sdk-darwin-arm64": ["@anthropic-ai/claude-agent-sdk-darwin-arm64@0.3.170", "", { "os": "darwin", "cpu": "arm64" }, "sha512-rwfgArIa5WI0QPNqFsRBgvtSI0mrtpynUm0oK6+l6/KX4hcgnYGEzciZR1bOeD9/7sSZlTdIgt+T9alKeZmXcg=="],
+    "@anthropic-ai/claude-agent-sdk-darwin-arm64": ["@anthropic-ai/claude-agent-sdk-darwin-arm64@0.3.197", "", { "os": "darwin", "cpu": "arm64" }, "sha512-jC6WvH5Hr6APTfbMjo4nC6LlyMMqbpCMwiHXIw7/AsQXIHQhZ+cRRMesQlV6UFI1l3O53gLZHzsG9cXwfrPHKw=="],

-    "@anthropic-ai/claude-agent-sdk-darwin-x64": ["@anthropic-ai/claude-agent-sdk-darwin-x64@0.3.170", "", { "os": "darwin", "cpu": "x64" }, "sha512-0e58h8UQMtsQxLGIv9r4foxfBFWKZ7NeDtoplLhuD7EwQonehomw1sBXCch77t/IfUS+q5vQ5zv+fOGmap5nLQ=="],
+    "@anthropic-ai/claude-agent-sdk-darwin-x64": ["@anthropic-ai/claude-agent-sdk-darwin-x64@0.3.197", "", { "os": "darwin", "cpu": "x64" }, "sha512-ZQNvGkMrTyatBlHTIQ4w2i2aLBuvq355UP/FDLnVXIH8l23RsL1x/0w9P+dqB7EmY9OZi/cPxSrpskpo+dZWLA=="],

-    "@anthropic-ai/claude-agent-sdk-linux-arm64": ["@anthropic-ai/claude-agent-sdk-linux-arm64@0.3.170", "", { "os": "linux", "cpu": "arm64" }, "sha512-gLbaFqcGppFJQd4DLNV4IXoeahejT/p2/M8bSSvRDbla9GOsBr1AxV5XLRyBn1e7xFGozZIAIQr3+1chp7NJgQ=="],
+    "@anthropic-ai/claude-agent-sdk-linux-arm64": ["@anthropic-ai/claude-agent-sdk-linux-arm64@0.3.197", "", { "os": "linux", "cpu": "arm64" }, "sha512-pWhQgCtAft4EGM4Zn24HRad1a/k2u6oA+2uM/KCdjehfKtooDiHfMNd1yzXY/n9AEBWP0RHB2Vz3mJ30X2pVAg=="],

-    "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": ["@anthropic-ai/claude-agent-sdk-linux-arm64-musl@0.3.170", "", { "os": "linux", "cpu": "arm64" }, "sha512-SRYfQcsXlOq+CD/FqkQBTSHbaD++w73GnnO+NUV9adLYrca3kfetRwWT1iguY1cNS0l34dCR3rlzCPq78vg1Jg=="],
+    "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": ["@anthropic-ai/claude-agent-sdk-linux-arm64-musl@0.3.197", "", { "os": "linux", "cpu": "arm64" }, "sha512-VuIGXsLGK/aqSQ0tTBqqPVNzjefWS5SWnK8mlYyQitT4s5UDzHXJm0UZBTGxRtlcS0e2+QAHKwbGBCq1ZKSXjg=="],

-    "@anthropic-ai/claude-agent-sdk-linux-x64": ["@anthropic-ai/claude-agent-sdk-linux-x64@0.3.170", "", { "os": "linux", "cpu": "x64" }, "sha512-Xl/m7TaSC3T5IDBdHrZQ9fCQYyDmPELN34CL+MoyPIf7uSmuZnjE9fUOqDh2Rv26JxWssi1M6X+BBvVuKd6Cpg=="],
+    "@anthropic-ai/claude-agent-sdk-linux-x64": ["@anthropic-ai/claude-agent-sdk-linux-x64@0.3.197", "", { "os": "linux", "cpu": "x64" }, "sha512-AUccrbdcv4Hy/GteP/gYLjG/zDP+fe2BFtDMctEfRFVz40DazYDcOyW1+nIgSTQtxf5jSTAVVf3cNuXB2CZwlw=="],

-    "@anthropic-ai/claude-agent-sdk-linux-x64-musl": ["@anthropic-ai/claude-agent-sdk-linux-x64-musl@0.3.170", "", { "os": "linux", "cpu": "x64" }, "sha512-m4+I0qBEk7cxRKS+pL+eoWXbXTFOAo83fQ0tQvap4z/mDMm06IWJtEPoYTaMBwsp32GJWLkHWKbZSBCHZnp2DQ=="],
+    "@anthropic-ai/claude-agent-sdk-linux-x64-musl": ["@anthropic-ai/claude-agent-sdk-linux-x64-musl@0.3.197", "", { "os": "linux", "cpu": "x64" }, "sha512-3Tuy7XhD4UIKE4A4RPmKJcbL7Q/3dcB1hEWQt2lKP7c/DlixeEv+tRzvpnFZKhFX2hy0tkBk3QjkozSAacMC/w=="],

-    "@anthropic-ai/claude-agent-sdk-win32-arm64": ["@anthropic-ai/claude-agent-sdk-win32-arm64@0.3.170", "", { "os": "win32", "cpu": "arm64" }, "sha512-IG+8isJNNJKbnnhO7m+PGhfVCg+XoQ/MDxGde5eigFI0WsEfitjuWSWwx82bT9ghxI1aa6qNvI+UPgPcZuo5Fg=="],
+    "@anthropic-ai/claude-agent-sdk-win32-arm64": ["@anthropic-ai/claude-agent-sdk-win32-arm64@0.3.197", "", { "os": "win32", "cpu": "arm64" }, "sha512-Wx8uiAKBenDuL8lWQmrqnX5ppljaH5unQ9cKiCz2/9Kgf09dgnrwbX8n/FhndCZR8PmYw539eWwYVrSVc/bl6w=="],

-    "@anthropic-ai/claude-agent-sdk-win32-x64": ["@anthropic-ai/claude-agent-sdk-win32-x64@0.3.170", "", { "os": "win32", "cpu": "x64" }, "sha512-7cuqSKbHVItPGVwRbd3A0BEJwcNtc7Fhoh6qHN4C6yrmjSrvdYYx3MLvq/VI768/RoG7mAMDxb+j7WfEfoP9BA=="],
+    "@anthropic-ai/claude-agent-sdk-win32-x64": ["@anthropic-ai/claude-agent-sdk-win32-x64@0.3.197", "", { "os": "win32", "cpu": "x64" }, "sha512-ZXJO/VvR3SI4G0gwthWeFXWdHB5RXPu3rtfGRcKZ/YgtDeW17rQ+LZIJTk2ywzbLb8EvlghR5JPgn293hC179Q=="],

-    "@anthropic-ai/sdk": ["@anthropic-ai/sdk@0.100.0", "", { "dependencies": { "json-schema-to-ts": "^3.1.1", "standardwebhooks": "^1.0.0" }, "peerDependencies": { "zod": "^3.25.0 || ^4.0.0" }, "optionalPeers": ["zod"], "bin": { "anthropic-ai-sdk": "bin/cli" } }, "sha512-cAm3aXm6qAiHIvHxyIIGd6tVmsD2gDqlc2h0R20ijNUzGgVnIN822bit4mKbF6CkuV7qIrLQIPoAepHEpanrQQ=="],
+    "@anthropic-ai/sdk": ["@anthropic-ai/sdk@0.108.0", "", { "dependencies": { "json-schema-to-ts": "^3.1.1", "standardwebhooks": "^1.0.0" }, "peerDependencies": { "zod": "^3.25.0 || ^4.0.0" }, "optionalPeers": ["zod"], "bin": { "anthropic-ai-sdk": "bin/cli" } }, "sha512-XBnl7Nszpbzg0aLnOCmdBi0bOU5goAsQ/L+NPNiuUPowDj8Mbzx0vlIIc1M79BjIvmw5nUM5G3jbrCBStT/0fQ=="],

    "@babel/runtime": ["@babel/runtime@7.29.2", "", {}, "sha512-JiDShH45zKHWyGe4ZNVRrCjBz8Nh9TMmZG1kh4QTK8hCBTWBi8Da+i7s1fJw7/lYpM4ccepSNfqzZ/QvABBi5g=="],

@@ -9,8 +9,8 @@
    "test": "bun test"
  },
  "dependencies": {
-    "@anthropic-ai/claude-agent-sdk": "^0.3.170",
-    "@anthropic-ai/sdk": "^0.100.0",
+    "@anthropic-ai/claude-agent-sdk": "^0.3.197",
+    "@anthropic-ai/sdk": "^0.108.0",
    "@modelcontextprotocol/sdk": "^1.29.0",
    "cron-parser": "^5.0.0",
    "zod": "^4.0.0"
@@ -1,5 +1,5 @@
 [
  { "name": "vercel", "version": "52.2.1" },
  { "name": "agent-browser", "version": "0.27.1", "onlyBuilt": true },
-  { "name": "@anthropic-ai/claude-code", "version": "2.1.170", "onlyBuilt": true }
+  { "name": "@anthropic-ai/claude-code", "version": "2.1.197", "onlyBuilt": true }
 ]
@@ -0,0 +1,177 @@
+# Agent Templates
+
+A **template** is a reusable folder you stamp into a working agent group: it
+carries the agent's standing instructions, its MCP tool servers, and its skills,
+but **no secrets and no provider**. Point `ncl` (or the setup wizard) at one and
+you get a configured agent in seconds; you choose the runtime/provider
+separately.
+
+Templates are purely additive: no DB migration, no new dependency. **At runtime,
+templates are resolved only from a local directory**: `templates/` at the
+project root by default (committed but shipped empty), or whatever
+`NANOCLAW_TEMPLATES_DIR` points at (a local path only). The setup wizard can also
+discover templates from the public registry
+([`nanocoai/nanoclaw-templates`](https://github.com/nanocoai/nanoclaw-templates))
+and copy a chosen one into your local `templates/` before stamping.
+
+## Using a template
+
+**During install.** `bash nanoclaw.sh` opens the setup wizard. Choose **Template
+setup**, then either **NanoClaw template library** (clones the public registry,
+copies the template you pick into your local `templates/`) or **Local templates**
+(lists what's already in `templates/`). The normal auth step then picks the
+runtime, and the wizard stamps and wires your first agent.
+
+**Anytime, via the CLI:**
+
+```bash
+ncl groups create --template sales/sdr --name "SDR Agent"
+```
+
+This stamps the group but does **not** wire it to a channel. Run
+`/manage-channels` (or `ncl wirings create`) afterward, exactly as for a
+hand-built group.
+
+### The template ref
+
+`--template <ref>` is a path **relative to the local templates directory**
+(`templates/` by default, or `NANOCLAW_TEMPLATES_DIR`). Refs are multi-segment,
+e.g. `sales/sdr` → `templates/sales/sdr`.
+
+For safety the ref must stay inside the templates directory: absolute paths, a
+leading `~`, and `../` escapes are rejected. There is no `--source`, no git URL,
+and no remote fetch at `ncl` time. Populate `templates/` first (by hand, or via
+the setup wizard's library option), then stamp.
+
+`NANOCLAW_TEMPLATES_DIR` may point the library at another **local** directory; it
+is never a URL and never changes at runtime.
+
+## What's in a template
+
+The full authoring reference lives in the
+[templates repo README](https://github.com/nanocoai/nanoclaw-templates#anatomy-of-a-template).
+The short version: only `context/instructions.md` is required; everything else
+is optional and defaults sensibly:
+
+```
+<template>/
+├── context/
+│   ├── instructions.md        # REQUIRED: the agent's standing persona; marks the folder as a template
+│   └── additional_context/    # optional: extra .md files, referenced from instructions.md by relative path
+│       └── *.md
+├── .mcp.json             # optional: MCP servers (command + args), NO secrets
+├── skills/<name>/        # optional: one folder per skill (SKILL.md + any references/), copied whole
+└── README.md             # recommended: per-template docs
+```
+
+| Path | Loaded as | Required |
+|------|-----------|----------|
+| `context/instructions.md` | The agent's persona, prepended to its `CLAUDE.md`/`AGENTS.md` every spawn (system-prompt tier, any provider) | **Yes** |
+| `context/**/*.md` (others) | Extra context, copied into the agent's workspace with the same layout relative to `instructions.md` | No |
+| `.mcp.json` → `mcpServers` | MCP tool servers (written verbatim to container config) | No |
+| `skills/<name>/` | A skill, auto-triggered by its `description` | No |
+
+Notes:
+
+- **No provider, model, effort, or packages in a template.** Those are set on
+  the agent later via `ncl groups config update`. The runtime defaults to the
+  install's configured provider.
+- **Keep `instructions.md` focused (under ~200 lines).** It's always in the
+  agent's prompt, and some providers cap that doc (Codex ~32 KB), so an over-long
+  persona gets truncated. Put bulk material in `skills/` or extra context files instead.
+- Skills are copied into the agent's own skills overlay, keyed to that group,
+  never shared across groups.
+
+### Referencing extra context files
+
+Extra `.md` files under `context/` (by convention in an `additional_context/`
+subfolder) are copied into the agent's workspace preserving their position
+relative to `instructions.md` — a template file at
+`context/additional_context/pricing.md` is readable by the agent as
+`additional_context/pricing.md`, the same relative path you'd use from
+`instructions.md` itself. Nothing is injected automatically: the agent only
+reads an extra file if `instructions.md` points to it, so reference every file
+you ship.
+
+```markdown
+Pricing rules live in `additional_context/pricing.md`. Read it before quoting a price.
+```
+
+Context files are copied when you stamp, so files added to the template later
+won't reach an already-created agent. Re-stamp the same name to update it.
+
+## MCP servers and credentials
+
+**Templates declare MCP servers, not secrets.** `.mcp.json` carries `command` +
+`args` only:
+
+```json
+{
+  "mcpServers": {
+    "hubspot": { "command": "npx", "args": ["-y", "@hubspot/mcp-server"] },
+    "exa":     { "command": "npx", "args": ["-y", "exa-mcp-server"] }
+  }
+}
+```
+
+Credentials are held by the **credentials proxy** and injected into outbound
+HTTPS calls at the proxy boundary, matched by API host, at request time. The key
+never sits in `.mcp.json`, the container env, or chat context. See
+[the credentials proxy section in CLAUDE.md](../CLAUDE.md#secrets--credentials--onecli)
+for the model.
+
+Two ways a credential gets connected:
+
+1. **Up front.** Register the secret with the credentials proxy (its web UI or
+   CLI), matched to the service's API host (e.g. `api.example.com`). Matching
+   credentials are injected automatically, so usually nothing else is needed.
+2. **On demand (the common path).** Don't set anything up first. The first time
+   the agent calls a service with no credential, the API returns **401/403** and
+   the agent replies with a prefilled connect link for that host. The user opens
+   it, pastes the key, and asks the agent to retry. The key lands in the
+   credentials proxy, which injects it on every later call.
+
+### MCP servers that require an env var to boot
+
+Some MCP servers refuse to start unless an env var is *present*, even though the
+real credential should come from the credentials proxy, not the env. Because `.mcp.json`'s `env`
+block passes through verbatim to the agent's container config, put a **placeholder
+value** there to satisfy the boot check:
+
+```json
+{
+  "mcpServers": {
+    "acme": {
+      "command": "npx",
+      "args": ["-y", "@acme/mcp-server"],
+      "env": { "ACME_API_KEY": "placeholder" }
+    }
+  }
+}
+```
+
+The server starts; its real outbound calls are still authenticated by the
+credentials proxy. **Never put a real key in `env`**: a placeholder only, and only when
+the server won't boot without one.
+
+### Approval-gating sensitive actions
+
+The credentials proxy can *hold* a credentialed outbound request and require a
+human to approve it before it leaves the proxy: enforcement the agent can't talk
+around. This is matched on the outbound HTTP request (host + method + path),
+configured on the credentials proxy, and answered by NanoClaw (it DMs an approver). The host side is
+already wired; see
+[the credentialed-approval flow in CLAUDE.md](../CLAUDE.md#requiring-approval-for-credential-use)
+and the [`sales/sdr` template README](https://github.com/nanocoai/nanoclaw-templates/blob/main/sales/sdr/README.md)
+for a worked example.
+
+## Contributing a template
+
+Templates ship in the separate
+[`nanocoai/nanoclaw-templates`](https://github.com/nanocoai/nanoclaw-templates)
+repo, not this one. To add one: fork that repo, drop a folder at
+`<category>/<template>/` with at least `context/instructions.md`, test it end to
+end (copy it under `templates/` and run
+`ncl groups create --template <category>/<template> --name Test`), confirm
+no secrets are committed, and open a PR. The repo's README has the full anatomy,
+category conventions, and checklist.
@@ -1,6 +1,6 @@
 {
  "name": "nanoclaw",
-  "version": "2.1.19",
+  "version": "2.1.24",
  "description": "Personal Claude assistant. Lightweight, secure, customizable.",
  "type": "module",
  "packageManager": "pnpm@10.33.0",
@@ -1,5 +1,5 @@
-<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="90" height="20" role="img" aria-label="199k tokens, 100% of context window">
-  <title>199k tokens, 100% of context window</title>
+<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="90" height="20" role="img" aria-label="204k tokens, 102% of context window">
+  <title>204k tokens, 102% of context window</title>
  <linearGradient id="s" x2="0" y2="100%">
    <stop offset="0" stop-color="#bbb" stop-opacity=".1"/>
    <stop offset="1" stop-opacity=".1"/>
@@ -15,8 +15,8 @@
      <g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" font-size="11">
        <text aria-hidden="true" x="26" y="15" fill="#010101" fill-opacity=".3">tokens</text>
        <text x="26" y="14">tokens</text>
-        <text aria-hidden="true" x="71" y="15" fill="#010101" fill-opacity=".3">199k</text>
-        <text x="71" y="14">199k</text>
+        <text aria-hidden="true" x="71" y="15" fill="#010101" fill-opacity=".3">204k</text>
+        <text x="71" y="14">204k</text>
      </g>
    </g>
  </a>
@@ -1,10 +1,11 @@
 #!/usr/bin/env bash
 #
-# Install the Slack adapter, persist SLACK_BOT_TOKEN + SLACK_SIGNING_SECRET to
+# Install the Slack adapter, persist SLACK_BOT_TOKEN plus the mode-specific
+# secret (SLACK_APP_TOKEN for Socket Mode, SLACK_SIGNING_SECRET for webhook) to
 # .env + data/env/env, and restart the service. Non-interactive — the
 # operator-facing app creation walkthrough + credential paste live in
 # setup/channels/slack.ts. Credentials come in via env vars:
-# SLACK_BOT_TOKEN, SLACK_SIGNING_SECRET.
+# SLACK_BOT_TOKEN, and SLACK_APP_TOKEN and/or SLACK_SIGNING_SECRET.
 #
 # Emits exactly one status block on stdout (ADD_SLACK) at the end. All chatty
 # progress messages go to stderr so setup:auto's raw-log capture sees the full
@@ -41,8 +42,10 @@ if [ -z "${SLACK_BOT_TOKEN:-}" ]; then
  emit_status failed "SLACK_BOT_TOKEN env var not set"
  exit 1
 fi
-if [ -z "${SLACK_SIGNING_SECRET:-}" ]; then
-  emit_status failed "SLACK_SIGNING_SECRET env var not set"
+# Socket Mode authenticates with SLACK_APP_TOKEN; webhook mode with
+# SLACK_SIGNING_SECRET. Require at least one.
+if [ -z "${SLACK_APP_TOKEN:-}" ] && [ -z "${SLACK_SIGNING_SECRET:-}" ]; then
+  emit_status failed "Set SLACK_APP_TOKEN (Socket Mode) or SLACK_SIGNING_SECRET (webhook)"
  exit 1
 fi

@@ -98,7 +101,12 @@ upsert_env() {
  fi
 }
 upsert_env SLACK_BOT_TOKEN "$SLACK_BOT_TOKEN"
-upsert_env SLACK_SIGNING_SECRET "$SLACK_SIGNING_SECRET"
+if [ -n "${SLACK_APP_TOKEN:-}" ]; then
+  upsert_env SLACK_APP_TOKEN "$SLACK_APP_TOKEN"
+fi
+if [ -n "${SLACK_SIGNING_SECRET:-}" ]; then
+  upsert_env SLACK_SIGNING_SECRET "$SLACK_SIGNING_SECRET"
+fi

 # Container reads from data/env/env (the host mounts it).
 mkdir -p data/env
@@ -4,15 +4,18 @@
 * `runSlackChannel(displayName)` owns the full branch from creating a
 * Slack app through the welcome DM:
 *
- *   1. Walk through creating a Slack app (api.slack.com/apps) — scopes,
- *      event subscriptions, and signing secret
- *   2. Paste the bot token + signing secret (clack password prompts)
- *   3. Validate via auth.test → resolves workspace + bot identity
- *   4. Install the adapter (setup/add-slack.sh, non-interactive)
- *   5. Ask for the operator's Slack user ID
- *   6. conversations.open to get the DM channel ID
- *   7. Ask for the messaging-agent name (defaulting to "Nano")
- *   8. Wire the agent via scripts/init-first-agent.ts
+ *   1. Ask the delivery mode: Socket Mode (outbound WebSocket, no public
+ *      URL) or a public webhook
+ *   2. Walk through creating a Slack app (api.slack.com/apps) — scopes,
+ *      events, and the mode-specific credential (app-level token for
+ *      Socket Mode, signing secret for webhook)
+ *   3. Paste the bot token + that credential (clack password prompts)
+ *   4. Validate via auth.test → resolves workspace + bot identity
+ *   5. Install the adapter (setup/add-slack.sh, non-interactive)
+ *   6. Ask for the operator's Slack user ID
+ *   7. conversations.open to get the DM channel ID
+ *   8. Ask for the messaging-agent name (defaulting to "Nano")
+ *   9. Wire the agent via scripts/init-first-agent.ts
 *
 * The welcome DM is sent via outbound delivery (chat.postMessage), which
 * works without Event Subscriptions being configured. The user sees the
@@ -45,14 +48,26 @@ interface WorkspaceInfo {
  botUserId: string;
 }

+// Socket Mode (SLACK_APP_TOKEN, xapp-…) needs no public URL; webhook mode
+// (SLACK_SIGNING_SECRET) needs a public Request URL. The adapter picks the mode
+// purely from SLACK_APP_TOKEN's presence — this choice just decides which
+// credential to collect and which post-install guidance to show.
+type SlackMode = 'socket' | 'webhook';
+
 export async function runSlackChannel(displayName: string): Promise<ChannelFlowResult> {
-  const intro = await walkThroughAppCreation();
+  const mode = await askSlackMode();
+  const intro = await walkThroughAppCreation(mode);
  if (intro === 'back') return BACK_TO_CHANNEL_SELECTION;

  const token = await collectBotToken();
-  const signingSecret = await collectSigningSecret();
+  const appToken = mode === 'socket' ? await collectAppToken() : undefined;
+  const signingSecret = mode === 'webhook' ? await collectSigningSecret() : undefined;
  const info = await validateSlackToken(token);

+  const env: Record<string, string> = { SLACK_BOT_TOKEN: token };
+  if (appToken) env.SLACK_APP_TOKEN = appToken;
+  if (signingSecret) env.SLACK_SIGNING_SECRET = signingSecret;
+
  const install = await runQuietChild(
    'slack-install',
    'bash',
@@ -62,11 +77,9 @@ export async function runSlackChannel(displayName: string): Promise<ChannelFlowR
      done: 'Slack adapter installed.',
    },
    {
-      env: {
-        SLACK_BOT_TOKEN: token,
-        SLACK_SIGNING_SECRET: signingSecret,
-      },
+      env,
      extraFields: {
+        MODE: mode,
        BOT_NAME: info.botName,
        TEAM_NAME: info.teamName,
        TEAM_ID: info.teamId,
@@ -122,10 +135,45 @@ export async function runSlackChannel(displayName: string): Promise<ChannelFlowR
    );
  }

-  showPostInstallChecklist(info);
+  showPostInstallChecklist(info, mode);
 }

-async function walkThroughAppCreation(): Promise<'continue' | 'back'> {
+async function askSlackMode(): Promise<SlackMode> {
+  const choice = ensureAnswer(
+    await brightSelect<SlackMode>({
+      message: 'How should Slack deliver events to NanoClaw?',
+      initialValue: 'socket',
+      options: [
+        {
+          value: 'socket',
+          label: 'Socket Mode',
+          hint: 'no public URL — recommended for local or behind NAT',
+        },
+        {
+          value: 'webhook',
+          label: 'Public webhook',
+          hint: 'needs a public HTTPS Request URL',
+        },
+      ],
+    }),
+  );
+  setupLog.userInput('slack_mode', String(choice));
+  return choice;
+}
+
+async function walkThroughAppCreation(mode: SlackMode): Promise<'continue' | 'back'> {
+  const credSteps =
+    mode === 'socket'
+      ? [
+          '  4. Basic Information → App-Level Tokens → "Generate Token and',
+          '     Scopes" → add the connections:write scope → copy it (xapp-…)',
+          '  5. Socket Mode → toggle "Enable Socket Mode" on',
+          '  6. Install to Workspace → copy the "Bot User OAuth Token" (xoxb-…)',
+        ]
+      : [
+          '  4. Basic Information → copy the "Signing Secret"',
+          '  5. Install to Workspace → copy the "Bot User OAuth Token" (xoxb-…)',
+        ];
  // Bright-white ANSI overrides the surrounding brand-cyan from `note()`'s
  // per-line formatter so the URL stands out against the rest of the body.
  const linkBlock = isHeadless()
@@ -149,8 +197,7 @@ async function walkThroughAppCreation(): Promise<'continue' | 'back'> {
      '     • files:read, files:write',
      '  3. App Home → enable "Messages Tab" and "Allow users to send',
      '     slash commands and messages from the messages tab"',
-      '  4. Basic Information → copy the "Signing Secret"',
-      '  5. Install to Workspace → copy the "Bot User OAuth Token" (xoxb-…)',
+      ...credSteps,
    ].join('\n'),
    'Create a Slack app',
  );
@@ -171,7 +218,10 @@ async function walkThroughAppCreation(): Promise<'continue' | 'back'> {

  ensureAnswer(
    await p.confirm({
-      message: 'Got your bot token and signing secret?',
+      message:
+        mode === 'socket'
+          ? 'Got your bot token and app-level token?'
+          : 'Got your bot token and signing secret?',
      initialValue: true,
    }),
  );
@@ -249,6 +299,40 @@ async function collectSigningSecret(): Promise<string> {
  return secret;
 }

+async function collectAppToken(): Promise<string> {
+  const existing = readEnvKey('SLACK_APP_TOKEN');
+  if (existing && existing.startsWith('xapp-') && existing.length >= 24) {
+    const reuse = ensureAnswer(await p.confirm({
+      message: `Found an existing Slack app-level token (${existing.slice(0, 10)}…). Use it?`,
+      initialValue: true,
+    }));
+    if (reuse) {
+      setupLog.userInput('slack_app_token', 'reused-existing');
+      return existing;
+    }
+  }
+
+  const answer = ensureAnswer(
+    await p.password({
+      message: 'Paste your Slack app-level token (Socket Mode)',
+      clearOnError: true,
+      validate: (v) => {
+        const t = (v ?? '').trim();
+        if (!t) return 'App-level token is required for Socket Mode';
+        if (!t.startsWith('xapp-')) return 'App-level tokens start with xapp-';
+        if (t.length < 24) return "That's shorter than a real Slack app-level token";
+        return undefined;
+      },
+    }),
+  );
+  const token = (answer as string).trim();
+  setupLog.userInput(
+    'slack_app_token',
+    `${token.slice(0, 10)}…${token.slice(-4)}`,
+  );
+  return token;
+}
+
 async function validateSlackToken(token: string): Promise<WorkspaceInfo> {
  const s = p.spinner();
  const start = Date.now();
@@ -416,7 +500,26 @@ async function resolveAgentName(): Promise<string> {
  return value;
 }

-function showPostInstallChecklist(info: WorkspaceInfo): void {
+function showPostInstallChecklist(info: WorkspaceInfo, mode: SlackMode): void {
+  if (mode === 'socket') {
+    note(
+      wrapForGutter(
+        [
+          `Your agent is wired to Slack and a welcome DM is on its way.`,
+          `Socket Mode is on — ${info.teamName} reaches NanoClaw over an outbound`,
+          `WebSocket, so there's no public URL to configure.`,
+          '',
+          '  • Just DM @' + info.botName + ' from Slack — replies flow straight away.',
+          '',
+          '  • Keep the NanoClaw host running to hold the socket open —',
+          '    Slack does not retry delivery while it is down.',
+        ].join('\n'),
+        6,
+      ),
+      'Finish setting up Slack',
+    );
+    return;
+  }
  note(
    wrapForGutter(
      [
@@ -43,7 +43,6 @@ interface V1Group {
  folder: string;
  trigger_pattern: string | null;
  requires_trigger: number | null;
-  is_main: number | null;
 }

 async function main(): Promise<void> {
@@ -65,7 +64,7 @@ async function main(): Promise<void> {
  // v1 schema varies — channel_name was a late addition. Query only the
  // columns we know exist in all v1 installs.
  const v1Groups = v1Db
-    .prepare('SELECT jid, name, folder, trigger_pattern, requires_trigger, is_main FROM registered_groups')
+    .prepare('SELECT jid, name, folder, trigger_pattern, requires_trigger FROM registered_groups')
    .all() as V1Group[];
  v1Db.close();

@@ -0,0 +1,93 @@
+import fs from 'fs';
+import path from 'path';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+const TEST_ROOT = '/tmp/nanoclaw-claude-md-compose-test';
+const GROUPS_DIR = path.join(TEST_ROOT, 'groups');
+
+vi.mock('./config.js', async (importOriginal) => ({
+  ...(await importOriginal<typeof import('./config.js')>()),
+  GROUPS_DIR: '/tmp/nanoclaw-claude-md-compose-test/groups',
+}));
+
+vi.mock('./log.js', () => ({
+  log: { debug: vi.fn(), info: vi.fn(), warn: vi.fn(), error: vi.fn(), fatal: vi.fn() },
+}));
+
+import { composeGroupClaudeMd } from './claude-md-compose.js';
+import { ensureContainerConfig } from './db/container-configs.js';
+import { closeDb, createAgentGroup, initTestDb, runMigrations } from './db/index.js';
+import { PERSONA_PREPEND_FILE } from './group-persona.js';
+import type { AgentGroup } from './types.js';
+
+function group(id: string, folder: string): AgentGroup {
+  return { id, name: folder, folder, agent_provider: null, created_at: new Date().toISOString() } as AgentGroup;
+}
+
+function seed(ag: AgentGroup): void {
+  createAgentGroup(ag);
+  ensureContainerConfig(ag.id);
+}
+
+function writePersona(folder: string, text: string): void {
+  const dir = path.join(GROUPS_DIR, folder);
+  fs.mkdirSync(dir, { recursive: true });
+  fs.writeFileSync(path.join(dir, PERSONA_PREPEND_FILE), text);
+}
+
+function importsOf(folder: string): string[] {
+  const md = fs.readFileSync(path.join(GROUPS_DIR, folder, 'CLAUDE.md'), 'utf-8');
+  return md.split('\n').filter((line) => line.startsWith('@'));
+}
+
+beforeEach(() => {
+  fs.rmSync(TEST_ROOT, { recursive: true, force: true });
+  fs.mkdirSync(TEST_ROOT, { recursive: true });
+  runMigrations(initTestDb());
+});
+
+afterEach(() => {
+  closeDb();
+  fs.rmSync(TEST_ROOT, { recursive: true, force: true });
+});
+
+describe('composeGroupClaudeMd persona prepend', () => {
+  it('imports the persona fragment FIRST, before the shared base', () => {
+    const ag = group('ag-persona', 'persona-group');
+    seed(ag);
+    writePersona(ag.folder, 'You are an SDR agent.\n');
+
+    composeGroupClaudeMd(ag);
+
+    const imports = importsOf(ag.folder);
+    expect(imports[0]).toBe('@./.claude-fragments/persona.md');
+    expect(imports[1]).toBe('@./.claude-shared.md');
+    expect(fs.readFileSync(path.join(GROUPS_DIR, ag.folder, '.claude-fragments', 'persona.md'), 'utf-8')).toBe(
+      'You are an SDR agent.',
+    );
+  });
+
+  it('keeps the persona across a second compose (not pruned)', () => {
+    const ag = group('ag-persona-2', 'persona-group-2');
+    seed(ag);
+    writePersona(ag.folder, 'persona body');
+
+    composeGroupClaudeMd(ag);
+    composeGroupClaudeMd(ag);
+
+    expect(fs.existsSync(path.join(GROUPS_DIR, ag.folder, '.claude-fragments', 'persona.md'))).toBe(true);
+    expect(importsOf(ag.folder)[0]).toBe('@./.claude-fragments/persona.md');
+  });
+
+  it('is inert when no persona file is present (non-template groups)', () => {
+    const ag = group('ag-no-persona', 'no-persona-group');
+    seed(ag);
+
+    composeGroupClaudeMd(ag);
+
+    const imports = importsOf(ag.folder);
+    expect(imports[0]).toBe('@./.claude-shared.md');
+    expect(imports).not.toContain('@./.claude-fragments/persona.md');
+    expect(fs.existsSync(path.join(GROUPS_DIR, ag.folder, '.claude-fragments', 'persona.md'))).toBe(false);
+  });
+});
@@ -20,9 +20,14 @@ import path from 'path';
 import { GROUPS_DIR } from './config.js';
 import type { McpServerConfig } from './container-config.js';
 import { getContainerConfig } from './db/container-configs.js';
+import { readGroupPersona } from './group-persona.js';
 import { log } from './log.js';
 import type { AgentGroup } from './types.js';

+// Fragment holding a template's persona prepend. Imported FIRST (before the
+// shared base) so the persona is the top of the composed system prompt.
+const PERSONA_FRAGMENT = 'persona.md';
+
 // Symlink targets are container paths — dangling on host (hence the readlink
 // dance instead of existsSync), valid inside the container via RO mounts.
 const SHARED_CLAUDE_MD_CONTAINER_PATH = '/app/CLAUDE.md';
@@ -106,6 +111,13 @@ export function composeGroupClaudeMd(group: AgentGroup): void {
    }
  }

+  // Template persona (if any) — inline so it survives the prune below; imported
+  // first (see the imports assembly) so it prepends the composed system prompt.
+  const persona = readGroupPersona(groupDir);
+  if (persona) {
+    desired.set(PERSONA_FRAGMENT, { type: 'inline', content: persona });
+  }
+
  // Reconcile: drop stale, write desired.
  for (const existing of fs.readdirSync(fragmentsDir)) {
    if (!desired.has(existing)) {
@@ -121,9 +133,14 @@ export function composeGroupClaudeMd(group: AgentGroup): void {
    }
  }

-  // Composed entry — imports only.
-  const imports = ['@./.claude-shared.md'];
-  for (const name of [...desired.keys()].sort()) {
+  // Composed entry — imports only. Persona first (top of the system prompt),
+  // then the shared base, then the remaining fragments sorted.
+  const imports: string[] = [];
+  if (desired.has(PERSONA_FRAGMENT)) {
+    imports.push(`@./.claude-fragments/${PERSONA_FRAGMENT}`);
+  }
+  imports.push('@./.claude-shared.md');
+  for (const name of [...desired.keys()].filter((n) => n !== PERSONA_FRAGMENT).sort()) {
    imports.push(`@./.claude-fragments/${name}`);
  }
  const body = [COMPOSED_HEADER, ...imports, ''].join('\n');
@@ -30,6 +30,8 @@ export interface ColumnDef {
  updatable?: boolean;
  /** Default value on create when not provided. */
  default?: unknown;
+  /** Default to another column's resolved value on create when not provided. */
+  defaultFrom?: string;
  /** Allowed values (shown in help). */
  enum?: string[];
 }
@@ -150,6 +152,8 @@ function genericCreate(def: ResourceDef) {
        throw new Error(`--${col.name.replace(/_/g, '-')} is required`);
      } else if (col.default !== undefined) {
        values[col.name] = col.default;
+      } else if (col.defaultFrom !== undefined && values[col.defaultFrom] !== undefined) {
+        values[col.name] = values[col.defaultFrom];
      }
    }

@@ -1,6 +1,9 @@
+import { randomUUID } from 'crypto';
+
 import type { McpServerConfig } from '../../container-config.js';
 import { buildAgentGroupImage, killContainer, wakeContainer } from '../../container-runner.js';
 import { restartAgentGroupContainers } from '../../container-restart.js';
+import { createAgentGroup } from '../../db/agent-groups.js';
 import { getDb, hasTable } from '../../db/connection.js';
 import { getSession } from '../../db/sessions.js';
 import { writeSessionMessage } from '../../session-manager.js';
@@ -9,7 +12,8 @@ import {
  updateContainerConfigScalars,
  updateContainerConfigJson,
 } from '../../db/container-configs.js';
-import type { ContainerConfigRow } from '../../types.js';
+import { createAgentFromTemplate } from '../../templates/create-agent.js';
+import type { AgentGroup, ContainerConfigRow } from '../../types.js';
 import { registerResource } from '../crud.js';

 /** Deserialize JSON columns for display. */
@@ -58,11 +62,37 @@ registerResource({
    },
    { name: 'created_at', type: 'string', description: 'Auto-set.', generated: true },
  ],
-  // `delete` is intentionally not in `operations` — the generic single-table
-  // DELETE violates FK constraints (see #2525). The cascading handler is
-  // provided as `customOperations.delete` below.
-  operations: { list: 'open', get: 'open', create: 'approval', update: 'approval' },
+  // `create` and `delete` are intentionally not in `operations` — create needs
+  // a `--template` branch (below); the generic single-table DELETE violates FK
+  // constraints (see #2525). Both are provided as `customOperations`.
+  operations: { list: 'open', get: 'open', update: 'approval' },
  customOperations: {
+    create: {
+      access: 'approval',
+      description:
+        'Create an agent group. With --template <ref>, stamp from a local template under templates/ ' +
+        '(MCP servers + instructions + skills); else insert a bare row (--name, --folder).',
+      handler: async (args) => {
+        if (args.template) {
+          return createAgentFromTemplate(String(args.template), {
+            name: args.name ? String(args.name) : undefined,
+          });
+        }
+        const name = args.name ? String(args.name) : '';
+        const folder = args.folder ? String(args.folder) : '';
+        if (!name) throw new Error('--name is required');
+        if (!folder) throw new Error('--folder is required');
+        const group: AgentGroup = {
+          id: randomUUID(),
+          name,
+          folder,
+          agent_provider: null,
+          created_at: new Date().toISOString(),
+        };
+        createAgentGroup(group);
+        return group;
+      },
+    },
    delete: {
      access: 'approval',
      description:
@@ -0,0 +1,75 @@
+/**
+ * Regression test: `ncl messaging-groups create` must satisfy the NOT NULL
+ * `instance` column without an operator-supplied `--instance`. The column has
+ * no CLI flag at the operator's altitude (the default instance IS the channel
+ * type), so the generic CRUD insert defaults it to `channel_type` — matching
+ * `createMessagingGroup`'s `instance ?? channel_type` fallback on the router
+ * path. Delete the `instance` column / `defaultFrom` wiring in
+ * `messaging-groups.ts` and this goes red: the insert fails the NOT NULL.
+ */
+import fs from 'fs';
+import { describe, expect, it, beforeEach, afterEach, vi } from 'vitest';
+
+vi.mock('../../container-runner.js', () => ({
+  wakeContainer: vi.fn().mockResolvedValue(undefined),
+  isContainerRunning: vi.fn().mockReturnValue(false),
+  getActiveContainerCount: vi.fn().mockReturnValue(0),
+  killContainer: vi.fn(),
+}));
+
+vi.mock('../../config.js', async () => {
+  const actual = await vi.importActual('../../config.js');
+  return { ...actual, DATA_DIR: '/tmp/nanoclaw-test-cli-msggroups' };
+});
+
+const TEST_DIR = '/tmp/nanoclaw-test-cli-msggroups';
+
+import { initTestDb, closeDb, runMigrations } from '../../db/index.js';
+import { getMessagingGroupByPlatform } from '../../db/messaging-groups.js';
+import { dispatch } from '../dispatch.js';
+// Side-effect import: registers the `messaging-groups-create` command.
+import './messaging-groups.js';
+
+describe('messaging-groups CLI create defaults instance to channel_type', () => {
+  beforeEach(() => {
+    if (fs.existsSync(TEST_DIR)) fs.rmSync(TEST_DIR, { recursive: true });
+    fs.mkdirSync(TEST_DIR, { recursive: true });
+    runMigrations(initTestDb());
+  });
+
+  afterEach(() => {
+    closeDb();
+    if (fs.existsSync(TEST_DIR)) fs.rmSync(TEST_DIR, { recursive: true });
+  });
+
+  it('create without --instance sets instance = channel_type', async () => {
+    // caller: 'host' is the post-approval re-entry path for create (approval op).
+    const resp = await dispatch(
+      {
+        id: 'req-1',
+        command: 'messaging-groups-create',
+        args: { channel_type: 'telegram', platform_id: '12345' },
+      },
+      { caller: 'host' },
+    );
+
+    expect(resp.ok).toBe(true);
+    const row = getMessagingGroupByPlatform('telegram', '12345');
+    expect(row).toBeDefined();
+    expect(row?.instance).toBe('telegram');
+  });
+
+  it('create with an explicit --instance keeps that value', async () => {
+    const resp = await dispatch(
+      {
+        id: 'req-2',
+        command: 'messaging-groups-create',
+        args: { channel_type: 'telegram', platform_id: '67890', instance: 'work' },
+      },
+      { caller: 'host' },
+    );
+
+    expect(resp.ok).toBe(true);
+    expect(getMessagingGroupByPlatform('telegram', '67890', 'work')?.instance).toBe('work');
+  });
+});
@@ -23,6 +23,14 @@ registerResource({
        'Platform-specific chat ID. Format varies: Telegram chat ID, Discord channel snowflake, Slack channel ID, phone number, email address.',
      required: true,
    },
+    {
+      name: 'instance',
+      type: 'string',
+      description:
+        'Adapter instance that owns this chat, when running N adapters of one channel type. Defaults to channel_type (the default instance) when omitted.',
+      defaultFrom: 'channel_type',
+      updatable: true,
+    },
    {
      name: 'name',
      type: 'string',
@@ -22,6 +22,12 @@ export const SENDER_ALLOWLIST_PATH = path.join(HOME_DIR, '.config', 'nanoclaw',
 export const STORE_DIR = path.resolve(PROJECT_ROOT, 'store');
 export const GROUPS_DIR = path.resolve(PROJECT_ROOT, 'groups');
 export const DATA_DIR = path.resolve(PROJECT_ROOT, 'data');
+// Local agent-template library. Committed but ships empty (+ README). Resolved
+// once at load. Override to another LOCAL path via NANOCLAW_TEMPLATES_DIR; never
+// a remote URL, never an ncl flag, never runtime-mutable.
+export const TEMPLATES_DIR = process.env.NANOCLAW_TEMPLATES_DIR
+  ? path.resolve(process.env.NANOCLAW_TEMPLATES_DIR)
+  : path.resolve(PROJECT_ROOT, 'templates');

 // Per-checkout image tag so two installs on the same host don't share
 // `nanoclaw-agent:latest` and clobber each other on rebuild.
@@ -185,6 +185,28 @@ export function updatePendingApprovalStatus(approvalId: string, status: PendingA
  getDb().prepare('UPDATE pending_approvals SET status = ? WHERE approval_id = ?').run(status, approvalId);
 }

+/**
+ * Park an approval in the "rejected, awaiting reason" hold: the admin clicked
+ * "Reject with reason…" and we're waiting for their one-line reply. `expiresAt`
+ * is the deadline after which the host sweep finalizes a plain reject (so a
+ * ghosted hold never strands the requesting agent). Reuses the otherwise-unused
+ * `expires_at` column on module-initiated rows.
+ */
+export function markApprovalAwaitingReason(approvalId: string, expiresAt: string): void {
+  getDb()
+    .prepare("UPDATE pending_approvals SET status = 'awaiting_reason', expires_at = ? WHERE approval_id = ?")
+    .run(expiresAt, approvalId);
+}
+
+/** Awaiting-reason approvals whose reply window has elapsed — the sweep's ghost set. */
+export function getExpiredAwaitingReasonApprovals(nowIso: string): PendingApproval[] {
+  return getDb()
+    .prepare(
+      "SELECT * FROM pending_approvals WHERE status = 'awaiting_reason' AND expires_at IS NOT NULL AND expires_at <= ?",
+    )
+    .all(nowIso) as PendingApproval[];
+}
+
 export function deletePendingApproval(approvalId: string): void {
  getDb().prepare('DELETE FROM pending_approvals WHERE approval_id = ?').run(approvalId);
 }
@@ -0,0 +1,32 @@
+import fs from 'fs';
+import path from 'path';
+import { afterEach, beforeEach, describe, expect, it } from 'vitest';
+
+import { PERSONA_PREPEND_FILE, readGroupPersona } from './group-persona.js';
+
+const TMP = '/tmp/nanoclaw-group-persona-test';
+
+beforeEach(() => {
+  fs.rmSync(TMP, { recursive: true, force: true });
+  fs.mkdirSync(TMP, { recursive: true });
+});
+
+afterEach(() => {
+  fs.rmSync(TMP, { recursive: true, force: true });
+});
+
+describe('readGroupPersona', () => {
+  it('returns null when the prepend file is absent', () => {
+    expect(readGroupPersona(TMP)).toBeNull();
+  });
+
+  it('returns null for an empty / whitespace-only file', () => {
+    fs.writeFileSync(path.join(TMP, PERSONA_PREPEND_FILE), '  \n\n');
+    expect(readGroupPersona(TMP)).toBeNull();
+  });
+
+  it('returns the trimmed content when present', () => {
+    fs.writeFileSync(path.join(TMP, PERSONA_PREPEND_FILE), '\nYou are an SDR agent.\n\n');
+    expect(readGroupPersona(TMP)).toBe('You are an SDR agent.');
+  });
+});
@@ -0,0 +1,30 @@
+/**
+ * Provider-neutral per-group persona ("instructions prepend").
+ *
+ * A template stamps its standing instructions here (src/templates/create-agent.ts).
+ * Each provider's project-doc composer inlines this content at the TOP of the
+ * doc it generates every spawn — `CLAUDE.md` (Claude, src/claude-md-compose.ts)
+ * or `AGENTS.md` (Codex, src/providers/codex-agents-md.ts on the providers
+ * branch) — so a template persona lands at system-prompt tier on every provider
+ * rather than in a recall-tier memory file.
+ *
+ * This module is the single owner of the filename + read semantics so the two
+ * composers (one on main, one on the providers donor branch) never hardcode the
+ * path independently. Absent file ⇒ null ⇒ no-op for non-template groups.
+ */
+import fs from 'fs';
+import path from 'path';
+
+/** Per-group host file holding the persona prepend. Never regenerated — persistent. */
+export const PERSONA_PREPEND_FILE = 'instructions.prepend.md';
+
+/**
+ * Read a group's persona prepend from its host dir, or null if absent/empty.
+ * `groupDir` is the per-group host directory (`GROUPS_DIR/<folder>`).
+ */
+export function readGroupPersona(groupDir: string): string | null {
+  const file = path.join(groupDir, PERSONA_PREPEND_FILE);
+  if (!fs.existsSync(file)) return null;
+  const content = fs.readFileSync(file, 'utf-8').trim();
+  return content.length > 0 ? content : null;
+}
@@ -0,0 +1,71 @@
+import fs from 'fs';
+import path from 'path';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+const TEST_ROOT = '/tmp/nanoclaw-group-skills-test';
+const DATA_DIR = path.join(TEST_ROOT, 'data');
+
+vi.mock('./config.js', async (importOriginal) => ({
+  ...(await importOriginal<typeof import('./config.js')>()),
+  DATA_DIR: '/tmp/nanoclaw-group-skills-test/data',
+}));
+
+import { materializeTemplateSkills } from './group-skills.js';
+
+function templateSkill(groupId: string, name: string, file: string, content: string): void {
+  const dir = path.join(DATA_DIR, 'v2-sessions', groupId, '.claude-shared', 'skills', name);
+  fs.mkdirSync(dir, { recursive: true });
+  fs.writeFileSync(path.join(dir, file), content);
+}
+
+beforeEach(() => {
+  fs.rmSync(TEST_ROOT, { recursive: true, force: true });
+  fs.mkdirSync(TEST_ROOT, { recursive: true });
+});
+
+afterEach(() => {
+  fs.rmSync(TEST_ROOT, { recursive: true, force: true });
+});
+
+describe('materializeTemplateSkills', () => {
+  it('copies real template-skill dirs into the provider skills dir', () => {
+    templateSkill('g1', 'widget', 'SKILL.md', 'body');
+    const dest = path.join(TEST_ROOT, 'grp1', '.agents', 'skills');
+
+    materializeTemplateSkills('g1', dest);
+
+    expect(fs.readFileSync(path.join(dest, 'widget', 'SKILL.md'), 'utf-8')).toBe('body');
+    expect(fs.lstatSync(path.join(dest, 'widget')).isSymbolicLink()).toBe(false);
+  });
+
+  it('is a no-op when the group has no template skills', () => {
+    const dest = path.join(TEST_ROOT, 'grp2', '.agents', 'skills');
+    materializeTemplateSkills('g2', dest);
+    expect(fs.existsSync(dest)).toBe(false);
+  });
+
+  it('overwrites its own skill dirs but leaves other destination entries intact', () => {
+    templateSkill('g3', 'widget', 'SKILL.md', 'new');
+    const dest = path.join(TEST_ROOT, 'grp3', '.agents', 'skills');
+    fs.mkdirSync(dest, { recursive: true });
+    // Stale copy of the same skill (should be refreshed) + a coexisting
+    // shared-skill symlink (must NOT be touched — it is provider-owned).
+    fs.mkdirSync(path.join(dest, 'widget'), { recursive: true });
+    fs.writeFileSync(path.join(dest, 'widget', 'SKILL.md'), 'old');
+    fs.symlinkSync('/app/skills/shared', path.join(dest, 'shared'));
+
+    materializeTemplateSkills('g3', dest);
+
+    expect(fs.readFileSync(path.join(dest, 'widget', 'SKILL.md'), 'utf-8')).toBe('new');
+    expect(fs.lstatSync(path.join(dest, 'shared')).isSymbolicLink()).toBe(true);
+  });
+
+  it('does not destroy skills when dest equals the source (Claude reads source directly)', () => {
+    templateSkill('g4', 'widget', 'SKILL.md', 'body');
+    const src = path.join(DATA_DIR, 'v2-sessions', 'g4', '.claude-shared', 'skills');
+
+    materializeTemplateSkills('g4', src);
+
+    expect(fs.existsSync(path.join(src, 'widget', 'SKILL.md'))).toBe(true);
+  });
+});
@@ -0,0 +1,52 @@
+/**
+ * Provider-agnostic template-skill materialization.
+ *
+ * A template stamps its skills as REAL directories into the group-private store
+ * `data/v2-sessions/<group-id>/.claude-shared/skills/<name>` (src/templates/create-agent.ts).
+ * Claude reads that store directly — it is mounted at `~/.claude/skills`, and
+ * real dirs survive the symlink-only skill-link prune. Every OTHER surfaces-owning
+ * provider (codex, opencode, pi, …) reads a DIFFERENT per-group skills directory,
+ * often READ-ONLY-mounted, so the skills must be copied there host-side, before
+ * the container starts.
+ *
+ * This is the single shared spot that does that copy. Each provider's host-side
+ * container contribution calls it once with its own skills dir (codex →
+ * `.agents/skills`; a future provider → whatever it reads). Adding a provider
+ * therefore adds one call, not a new mirror implementation. The copied dirs are
+ * real (not symlinks), so they survive providers' symlink-only prunes and persist
+ * across respawns.
+ *
+ * This module is a main-owned seam that provider payloads (on the `providers`
+ * donor branch) import — mirrors src/group-persona.ts.
+ */
+import fs from 'fs';
+import path from 'path';
+
+import { DATA_DIR } from './config.js';
+
+/** The group-private store templates stamp skills into (Claude's read plane). */
+function templateSkillsSource(agentGroupId: string): string {
+  return path.join(DATA_DIR, 'v2-sessions', agentGroupId, '.claude-shared', 'skills');
+}
+
+/**
+ * Copy a group's template skills into a provider's per-group skills directory.
+ * No-op if the group has no template skills, or if `destSkillsDir` IS the source
+ * (Claude, which reads the source directly — copying onto itself would delete it).
+ * Idempotent: overwrites each template skill so edits propagate on respawn. It
+ * manages only its own skill dirs — other entries in the destination (e.g. a
+ * provider's shared-skill symlinks) are left untouched.
+ */
+export function materializeTemplateSkills(agentGroupId: string, destSkillsDir: string): void {
+  const src = templateSkillsSource(agentGroupId);
+  if (!fs.existsSync(src)) return;
+  if (path.resolve(src) === path.resolve(destSkillsDir)) return;
+
+  fs.mkdirSync(destSkillsDir, { recursive: true });
+  for (const name of fs.readdirSync(src)) {
+    if (!fs.statSync(path.join(src, name)).isDirectory()) continue;
+    const dest = path.join(destSkillsDir, name);
+    fs.rmSync(dest, { recursive: true, force: true });
+    fs.cpSync(path.join(src, name), dest, { recursive: true });
+  }
+}
@@ -152,6 +152,18 @@ async function sweep(): Promise<void> {
    log.error('Host sweep error', { err });
  }

+  // Finalize any "Reject with reason…" holds whose reply window elapsed (admin
+  // ghosted, or the host restarted mid-capture). Central-DB scan, once per tick
+  // — not per session.
+  // MODULE-HOOK:approvals-reason-sweep:start
+  try {
+    const { sweepAwaitingReasonRejects } = await import('./modules/approvals/index.js');
+    await sweepAwaitingReasonRejects();
+  } catch (err) {
+    log.error('Reject-with-reason sweep failed', { err });
+  }
+  // MODULE-HOOK:approvals-reason-sweep:end
+
  setTimeout(sweep, SWEEP_INTERVAL_MS);
 }

@@ -0,0 +1,83 @@
+/**
+ * Shared containment guards for per-message inbox directories.
+ *
+ * Session dirs are mounted writable into agent containers, so a compromised
+ * agent can pre-place a symlink inside its own session dir and wait for the
+ * host to write through it — landing attacker-influenced bytes outside the
+ * sandbox (CWE-59). Both inbound paths that materialise files into a session's
+ * `inbox/<messageId>/` directory route through `ensureContainedInboxDir`:
+ *   - channel-inbound attachments (`extractAttachmentFiles` in session-manager)
+ *   - agent-to-agent forwarded files (`forwardAttachedFiles` in agent-route)
+ *
+ * Keeping the guard in one place means both paths defend identically; the fix
+ * for GHSA #2828 originally lived only in the A2A path and the channel path had
+ * the same gap (a symlinked `inbox` root was followed silently).
+ */
+import fs from 'fs';
+import path from 'path';
+
+import { log } from './log.js';
+
+/** True if `child` is `parent` itself or nested within it (no traversal/escape). */
+export function isPathInside(parent: string, child: string): boolean {
+  const relative = path.relative(parent, child);
+  return relative === '' || (!relative.startsWith('..') && !path.isAbsolute(relative));
+}
+
+/**
+ * Resolve and create `<inboxRoot>/<messageId>`, refusing pre-placed symlinks a
+ * compromised container could use to redirect host writes outside the session.
+ *
+ * Guards, in order:
+ *   1. lstat the inbox ROOT — reject if it is a symlink or a non-directory.
+ *      Without this, a symlinked `inbox` is silently followed by mkdir AND the
+ *      containment check in step 4 passes, because it compares against the
+ *      already-followed (escaped) root. This is the gap that affected the
+ *      channel-inbound path.
+ *   2. lstat the per-message subdir — reject a pre-placed symlink/non-dir.
+ *      lstat does not follow the final path component, so it sees the link
+ *      itself even when the link target does not exist.
+ *   3. mkdir the subdir (recursive).
+ *   4. realpath containment — the resolved subdir must stay within the resolved
+ *      inbox root (defence in depth; symlinks are already ruled out above).
+ *
+ * Returns the resolved, contained subdir path (write into it with an exclusive
+ * flag — `COPYFILE_EXCL` / `wx` — so a pre-existing symlinked *file* can't be
+ * followed either), or `null` if any guard tripped. On `null` the caller logs
+ * its own context and skips; `context` is merged into the warn logs here so
+ * each call site stays diagnosable.
+ */
+export function ensureContainedInboxDir(
+  inboxRoot: string,
+  messageId: string,
+  context: Record<string, unknown>,
+): string | null {
+  const inboxDir = path.join(inboxRoot, messageId);
+
+  for (const dir of [inboxRoot, inboxDir]) {
+    try {
+      const st = fs.lstatSync(dir);
+      if (st.isSymbolicLink() || !st.isDirectory()) {
+        log.warn('inbox-safety: rejecting unsafe inbox path', { ...context, dir });
+        return null;
+      }
+    } catch {
+      // Does not exist yet — fine, mkdir below creates it.
+    }
+  }
+
+  fs.mkdirSync(inboxDir, { recursive: true });
+
+  try {
+    const realInboxDir = fs.realpathSync(inboxDir);
+    const realInboxRoot = fs.realpathSync(inboxRoot);
+    if (!isPathInside(realInboxRoot, realInboxDir)) {
+      log.warn('inbox-safety: inbox dir escaped inbox root', { ...context, inboxDir });
+      return null;
+    }
+    return realInboxDir;
+  } catch (err) {
+    log.warn('inbox-safety: failed to resolve inbox dir', { ...context, inboxDir, err });
+    return null;
+  }
+}
@@ -3,7 +3,8 @@ import fs from 'fs';
 import path from 'path';
 import { describe, expect, it, beforeEach, afterEach, vi } from 'vitest';

-import { isSafeAttachmentName, routeAgentMessage } from './agent-route.js';
+import { forwardAttachedFiles, isSafeAttachmentName, routeAgentMessage } from './agent-route.js';
+import { log } from '../../log.js';
 import { createDestination } from './db/agent-destinations.js';
 import { initTestDb, closeDb, runMigrations, createAgentGroup } from '../../db/index.js';
 import { createSession, updateSession } from '../../db/sessions.js';
@@ -467,4 +468,129 @@ describe('routeAgentMessage return-path', () => {
    const parsed = JSON.parse(bRows[0].content);
    expect(parsed.attachments).toHaveLength(0);
  });
+
+  // #2828 — target-side symlink containment. A compromised target agent can
+  // write inside its own session dir; these tests prove it cannot redirect a
+  // forwarded attachment outside the session sandbox via a pre-placed symlink.
+
+  it('file forwarding (#2828): skips a symlinked target inbox dir, writes nothing outside', async () => {
+    const warnSpy = vi.spyOn(log, 'warn');
+    const canaryDir = path.join(TEST_DIR, 'canary-outside-inbox');
+    fs.mkdirSync(canaryDir, { recursive: true });
+
+    // Source has a real attachment to forward.
+    const outboxDir = path.join(sessionDir(A, S1.id), 'outbox', 'msg-evil-inbox');
+    fs.mkdirSync(outboxDir, { recursive: true });
+    fs.writeFileSync(path.join(outboxDir, 'pwn.txt'), 'attacker-bytes');
+
+    // Target pre-places its whole `inbox` as a symlink pointing outside.
+    const targetInbox = path.join(sessionDir(B, SB.id), 'inbox');
+    fs.rmSync(targetInbox, { recursive: true, force: true });
+    fs.symlinkSync(canaryDir, targetInbox);
+
+    await routeAgentMessage(
+      {
+        id: 'msg-evil-inbox',
+        platform_id: B,
+        content: JSON.stringify({ text: 'see attached', files: ['pwn.txt'] }),
+        in_reply_to: null,
+      },
+      S1,
+    );
+
+    // Message still routes — just with no attachments.
+    const bRows = readInbound(B, SB.id);
+    expect(bRows).toHaveLength(1);
+    expect(JSON.parse(bRows[0].content).attachments).toHaveLength(0);
+
+    // Nothing was written through the symlink to the canary location.
+    expect(fs.readdirSync(canaryDir)).toHaveLength(0);
+    expect(warnSpy).toHaveBeenCalled();
+    warnSpy.mockRestore();
+  });
+
+  it('file forwarding (#2828): skips a symlinked inbox/<msgId> subdir, writes nothing outside', async () => {
+    const warnSpy = vi.spyOn(log, 'warn');
+    const canaryDir = path.join(TEST_DIR, 'canary-outside-subdir');
+    fs.mkdirSync(canaryDir, { recursive: true });
+
+    const outboxDir = path.join(sessionDir(A, S1.id), 'outbox', 'msg-evil-subdir');
+    fs.mkdirSync(outboxDir, { recursive: true });
+    fs.writeFileSync(path.join(outboxDir, 'pwn.txt'), 'attacker-bytes');
+
+    // The forwarded a2a msg id generated inside routeAgentMessage is random, so
+    // a symlink can't be pre-placed at inbox/<that-id>. Drive forwardAttachedFiles
+    // directly with a fixed target message id and plant the symlink at that path.
+    const targetMsgId = 'evil-subdir-msg';
+    const realInbox = path.join(sessionDir(B, SB.id), 'inbox');
+    fs.mkdirSync(realInbox, { recursive: true });
+    fs.symlinkSync(canaryDir, path.join(realInbox, targetMsgId));
+
+    const attachments = forwardAttachedFiles(
+      { agentGroupId: A, sessionId: S1.id, messageId: 'msg-evil-subdir', filenames: ['pwn.txt'] },
+      { agentGroupId: B, sessionId: SB.id, messageId: targetMsgId },
+    );
+
+    expect(attachments).toHaveLength(0);
+    expect(fs.readdirSync(canaryDir)).toHaveLength(0);
+    expect(warnSpy).toHaveBeenCalled();
+    warnSpy.mockRestore();
+  });
+
+  it('file forwarding (#2828): refuses a pre-existing symlinked dst file (COPYFILE_EXCL)', async () => {
+    const warnSpy = vi.spyOn(log, 'warn');
+    const canaryFile = path.join(TEST_DIR, 'canary-dst-target.txt');
+    fs.writeFileSync(canaryFile, 'original-canary');
+
+    const outboxDir = path.join(sessionDir(A, S1.id), 'outbox', 'msg-evil-dst');
+    fs.mkdirSync(outboxDir, { recursive: true });
+    fs.writeFileSync(path.join(outboxDir, 'doc.txt'), 'attacker-bytes');
+
+    // inbox/<msgId>/ is a real dir, but contains a pre-placed symlink named
+    // exactly like the incoming attachment, pointing at the canary file.
+    // We can only do this once we know the a2a msg id, which is generated
+    // inside routeAgentMessage. So we instead drive forwardAttachedFiles
+    // directly with a fixed target message id.
+    const targetMsgId = 'fixed-evil-dst';
+    const realInboxSubdir = path.join(sessionDir(B, SB.id), 'inbox', targetMsgId);
+    fs.mkdirSync(realInboxSubdir, { recursive: true });
+    fs.symlinkSync(canaryFile, path.join(realInboxSubdir, 'doc.txt'));
+
+    const attachments = forwardAttachedFiles(
+      { agentGroupId: A, sessionId: S1.id, messageId: 'msg-evil-dst', filenames: ['doc.txt'] },
+      { agentGroupId: B, sessionId: SB.id, messageId: targetMsgId },
+    );
+
+    // The exclusive write failed → nothing forwarded.
+    expect(attachments).toHaveLength(0);
+    // Canary file untouched (symlink not followed/overwritten).
+    expect(fs.readFileSync(canaryFile, 'utf-8')).toBe('original-canary');
+    expect(warnSpy).toHaveBeenCalled();
+    warnSpy.mockRestore();
+  });
+
+  it('file forwarding (#2828 regression): a normal forward still works end-to-end', async () => {
+    const outboxDir = path.join(sessionDir(A, S1.id), 'outbox', 'msg-ok-file');
+    fs.mkdirSync(outboxDir, { recursive: true });
+    fs.writeFileSync(path.join(outboxDir, 'ok.txt'), 'legit-bytes');
+
+    await routeAgentMessage(
+      {
+        id: 'msg-ok-file',
+        platform_id: B,
+        content: JSON.stringify({ text: 'see attached', files: ['ok.txt'] }),
+        in_reply_to: null,
+      },
+      S1,
+    );
+
+    const bRows = readInbound(B, SB.id);
+    expect(bRows).toHaveLength(1);
+    const parsed = JSON.parse(bRows[0].content);
+    expect(parsed.attachments).toHaveLength(1);
+    expect(parsed.attachments[0].name).toBe('ok.txt');
+    const targetPath = path.join(sessionDir(B, SB.id), parsed.attachments[0].localPath);
+    expect(fs.existsSync(targetPath)).toBe(true);
+    expect(fs.readFileSync(targetPath, 'utf-8')).toBe('legit-bytes');
+  });
 });
@@ -22,6 +22,7 @@ import fs from 'fs';
 import path from 'path';

 import { isSafeAttachmentName } from '../../attachment-safety.js';
+import { ensureContainedInboxDir, isPathInside } from '../../inbox-safety.js';
 import { getAgentGroup } from '../../db/agent-groups.js';
 import { getInboundSourceSessionId, getMostRecentPeerSourceSessionId } from '../../db/session-db.js';
 import { getSession } from '../../db/sessions.js';
@@ -42,11 +43,6 @@ export interface ForwardedAttachment {
  localPath: string;
 }

-function isPathInside(parent: string, child: string): boolean {
-  const relative = path.relative(parent, child);
-  return relative === '' || (!relative.startsWith('..') && !path.isAbsolute(relative));
-}
-
 /**
 * Copy file attachments from the source agent's outbox into the target
 * agent's inbox. Returns attachments using the formatter's existing
@@ -98,8 +94,20 @@ export function forwardAttachedFiles(
    return [];
  }

-  const targetInboxDir = path.join(sessionDir(target.agentGroupId, target.sessionId), 'inbox', target.messageId);
-  fs.mkdirSync(targetInboxDir, { recursive: true });
+  // Target-side containment — shared with the channel-inbound path. A
+  // compromised target agent can write inside its own session dir, so it could
+  // pre-place `inbox` (or `inbox/<future-msgId>`) as a symlink pointing
+  // anywhere host-writable; ensureContainedInboxDir refuses the symlink before
+  // any copy lands outside the sandbox (#2828, CWE-59).
+  const inboxRoot = path.join(sessionDir(target.agentGroupId, target.sessionId), 'inbox');
+  const targetInboxDir = ensureContainedInboxDir(inboxRoot, target.messageId, {
+    targetGroup: target.agentGroupId,
+    targetSession: target.sessionId,
+    targetMsgId: target.messageId,
+  });
+  if (!targetInboxDir) {
+    return [];
+  }

  const attachments: ForwardedAttachment[] = [];
  for (const filename of source.filenames) {
@@ -137,7 +145,20 @@ export function forwardAttachedFiles(
      continue;
    }
    const dst = path.join(targetInboxDir, filename);
-    fs.copyFileSync(realSrc, dst);
+    try {
+      // COPYFILE_EXCL: fail with EEXIST rather than follow or overwrite a
+      // pre-placed symlink / existing file at dst — the host is the sole
+      // writer of these attachments.
+      fs.copyFileSync(realSrc, dst, fs.constants.COPYFILE_EXCL);
+    } catch (err) {
+      log.warn('agent-route: refusing to write target inbox file', {
+        sourceMsgId: source.messageId,
+        targetMsgId: target.messageId,
+        filename,
+        err,
+      });
+      continue;
+    }
    attachments.push({
      name: filename,
      filename,
@@ -278,7 +299,10 @@ function buildGateQuestion(sourceName: string, targetName: string, contentStr: s
  const body = text.length > GATE_CARD_BODY_MAX ? `${text.slice(0, GATE_CARD_BODY_MAX)}… (truncated)` : text;
  const lines = [`Agent "${sourceName}" wants to send a message to "${targetName}":`, '', body];
  if (files.length > 0) lines.push('', `Attachments: ${files.join(', ')}`);
-  lines.push('', 'Approve delivery?');
+  lines.push(
+    '',
+    `Approve, Reject, or "Reject with reason…" to decline and then type a short reason I'll relay to "${sourceName}".`,
+  );
  return lines.join('\n');
 }

@@ -0,0 +1,59 @@
+/**
+ * Shared "finalize a rejected approval" path.
+ *
+ * Three entry points land here so they relay one message and clean up
+ * identically:
+ *   1. The instant Reject button            (response-handler.ts)
+ *   2. A captured Reject-with-reason reply   (reason-capture.ts)
+ *   3. The host-sweep ghost finalizer        (reason-capture.ts, via host-sweep)
+ *
+ * Kept in its own leaf file so both response-handler.ts and reason-capture.ts
+ * can import it without an import cycle (finalize → primitive only).
+ */
+import { wakeContainer } from '../../container-runner.js';
+import { deletePendingApproval } from '../../db/sessions.js';
+import { log } from '../../log.js';
+import { writeSessionMessage } from '../../session-manager.js';
+import type { PendingApproval, Session } from '../../types.js';
+import { notifyApprovalResolved } from './primitive.js';
+
+/**
+ * Notify the requesting agent that its action was rejected, drop the pending
+ * row, fire approval-resolved callbacks, and wake the container.
+ *
+ * When `reason` is provided it's appended to the agent-facing note with generic
+ * attribution — the why, not the who (the rejecting admin may belong to a
+ * different owner than the requesting agent). Callers are responsible for
+ * clamping the reason length before passing it in.
+ */
+export async function finalizeReject(
+  approval: PendingApproval,
+  session: Session,
+  userId: string,
+  reason?: string,
+): Promise<void> {
+  const text = reason
+    ? `Your ${approval.action} request was rejected by admin: "${reason}"`
+    : `Your ${approval.action} request was rejected by admin.`;
+
+  writeSessionMessage(session.agent_group_id, session.id, {
+    id: `appr-note-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
+    kind: 'chat',
+    timestamp: new Date().toISOString(),
+    platformId: session.agent_group_id,
+    channelType: 'agent',
+    threadId: null,
+    content: JSON.stringify({ text, sender: 'system', senderId: 'system' }),
+  });
+
+  log.info('Approval rejected', {
+    approvalId: approval.approval_id,
+    action: approval.action,
+    userId,
+    withReason: reason !== undefined,
+  });
+
+  deletePendingApproval(approval.approval_id);
+  await notifyApprovalResolved({ approval, session, outcome: 'reject', userId });
+  await wakeContainer(session);
+}
@@ -8,10 +8,16 @@
 *   - A response handler that claims pending_approvals rows and dispatches
 *     to whatever module registered for the row's `action` string. Also
 *     resolves in-memory OneCLI credential approvals.
+ *   - A message-interceptor (via ./reason-capture.js) that captures an admin's
+ *     one-line reply after they click "Reject with reason…".
 *   - An adapter-ready callback that starts the OneCLI manual-approval handler
 *     once the delivery adapter is set.
 *   - A shutdown callback that stops the OneCLI handler cleanly.
 *
+ * Exposes `sweepAwaitingReasonRejects` for the host sweep to finalize ghosted
+ * reject-with-reason holds (re-exported here, which also loads reason-capture
+ * so its interceptor registers).
+ *
 * Self-mod flows (install_packages, add_mcp_server) moved out to
 * `src/modules/self-mod/` in PR #7 — they now register delivery actions
 * + approval handlers via this module's public API.
@@ -24,6 +30,9 @@ import { startOneCLIApprovalHandler, stopOneCLIApprovalHandler } from './onecli-
 // Public API re-exports so consumers import from the module root.
 export { requestApproval, registerApprovalHandler, notifyAgent } from './primitive.js';
 export type { ApprovalHandler, ApprovalHandlerContext, RequestApprovalOptions } from './primitive.js';
+// Host-sweep hook for ghosted "Reject with reason…" holds. The re-export also
+// loads reason-capture.js, registering its message-interceptor on import.
+export { sweepAwaitingReasonRejects } from './reason-capture.js';

 registerResponseHandler(handleApprovalsResponse);

@@ -32,10 +32,23 @@ import type { MessagingGroup, PendingApproval, Session } from '../../types.js';
 import { getAdminsOfAgentGroup, getGlobalAdmins, getOwners } from '../permissions/db/user-roles.js';
 import { ensureUserDm } from '../permissions/user-dm.js';

-/** Two-button approval UI — the only options the primitive supports today. */
+/**
+ * Card value for the "Reject with reason…" button. Selecting it doesn't
+ * finalize the reject — it holds the row and captures the approver's next DM
+ * as a one-line reason relayed to the requesting agent. See reason-capture.ts.
+ */
+export const REJECT_WITH_REASON_VALUE = 'reject_with_reason';
+
+/**
+ * Three-button approval UI. Plain Reject is the instant fast path; "Reject with
+ * reason…" opts into the reason-capture flow. Shared by every module approval
+ * (create_agent, install_packages, add_mcp_server); OneCLI credential cards
+ * keep their own two-button set in onecli-approvals.ts.
+ */
 const APPROVAL_OPTIONS: RawOption[] = [
  { label: 'Approve', selectedLabel: '✅ Approved', value: 'approve' },
  { label: 'Reject', selectedLabel: '❌ Rejected', value: 'reject' },
+  { label: 'Reject with reason…', selectedLabel: '📝 Rejected (awaiting reason)', value: REJECT_WITH_REASON_VALUE },
 ];

 // ── Approval handler registry ──
@@ -0,0 +1,279 @@
+/**
+ * "Reject with reason…" capture flow.
+ *
+ * Covers the three entry points end to end against the real central DB:
+ *   - arming (handleApprovalsResponse with the third option) holds the row and
+ *     prompts the admin instead of finalizing;
+ *   - the captured reply relays one combined message, clamped to 280 chars;
+ *   - the host sweep finalizes a ghosted hold as a plain reject.
+ *
+ * writeSessionMessage is mocked so the relayed agent-facing text can be read
+ * back directly; the delivery adapter is a fake that records prompt sends.
+ */
+import * as fs from 'fs';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+import type { InboundEvent } from '../../channels/adapter.js';
+import { initTestDb, closeDb, runMigrations } from '../../db/index.js';
+import { createAgentGroup } from '../../db/agent-groups.js';
+import { createMessagingGroup } from '../../db/messaging-groups.js';
+import {
+  createSession,
+  createPendingApproval,
+  deletePendingApproval,
+  getPendingApproval,
+  markApprovalAwaitingReason,
+} from '../../db/sessions.js';
+import { setDeliveryAdapter, type ChannelDeliveryAdapter } from '../../delivery.js';
+import { writeSessionMessage } from '../../session-manager.js';
+import { upsertUser } from '../permissions/db/users.js';
+import { upsertUserDm } from '../permissions/db/user-dms.js';
+import { grantRole } from '../permissions/db/user-roles.js';
+import { REJECT_WITH_REASON_VALUE } from './primitive.js';
+
+vi.mock('../../container-runner.js', () => ({
+  wakeContainer: vi.fn().mockResolvedValue(undefined),
+}));
+
+vi.mock('../../config.js', async () => {
+  const actual = await vi.importActual('../../config.js');
+  return { ...actual, DATA_DIR: '/tmp/nanoclaw-test-reject-reason' };
+});
+
+vi.mock('../../session-manager.js', async () => {
+  const actual = await vi.importActual<typeof import('../../session-manager.js')>('../../session-manager.js');
+  return { ...actual, writeSessionMessage: vi.fn() };
+});
+
+const TEST_DIR = '/tmp/nanoclaw-test-reject-reason';
+const DM_CHANNEL = 'slack';
+const DM_PLATFORM = 'D-admin-1';
+
+function now(): string {
+  return new Date().toISOString();
+}
+
+let delivered: Array<{ channelType: string; platformId: string; content: string }>;
+
+const fakeAdapter: ChannelDeliveryAdapter = {
+  async deliver(channelType, platformId, _threadId, _kind, content) {
+    delivered.push({ channelType, platformId, content });
+    return 'pm-1';
+  },
+};
+
+function seedApproval(approvalId: string, action = 'create_agent'): void {
+  createPendingApproval({
+    approval_id: approvalId,
+    session_id: 'sess-1',
+    request_id: approvalId,
+    action,
+    payload: JSON.stringify({ name: 'child' }),
+    created_at: now(),
+    title: 'Approval',
+    options_json: JSON.stringify([]),
+  });
+}
+
+function dmReply(text?: string): InboundEvent {
+  const content: Record<string, unknown> = { sender: 'admin-1', senderId: 'admin-1' };
+  if (text !== undefined) content.text = text;
+  return {
+    channelType: DM_CHANNEL,
+    platformId: DM_PLATFORM,
+    threadId: null,
+    message: { id: 'm-1', kind: 'chat', content: JSON.stringify(content), timestamp: now() },
+  };
+}
+
+/** Click the "Reject with reason…" button as the seeded admin. */
+async function clickRejectWithReason(approvalId: string): Promise<void> {
+  const { handleApprovalsResponse } = await import('./response-handler.js');
+  await handleApprovalsResponse({
+    questionId: approvalId,
+    value: REJECT_WITH_REASON_VALUE,
+    userId: 'admin-1',
+    channelType: DM_CHANNEL,
+    platformId: '', // not surfaced by the click payload — resolved via ensureUserDm
+    threadId: null,
+  });
+}
+
+/** The text of the most recent agent-facing note written via writeSessionMessage. */
+function lastRelayedText(): string | undefined {
+  const call = vi.mocked(writeSessionMessage).mock.calls.at(-1);
+  if (!call) return undefined;
+  return (JSON.parse(call[2].content) as { text: string }).text;
+}
+
+beforeEach(() => {
+  vi.clearAllMocks();
+  if (fs.existsSync(TEST_DIR)) fs.rmSync(TEST_DIR, { recursive: true, force: true });
+  fs.mkdirSync(TEST_DIR, { recursive: true });
+  const db = initTestDb();
+  runMigrations(db);
+  delivered = [];
+
+  createAgentGroup({ id: 'ag-1', name: 'Agent', folder: 'agent', agent_provider: null, created_at: now() });
+  createSession({
+    id: 'sess-1',
+    agent_group_id: 'ag-1',
+    messaging_group_id: null,
+    thread_id: null,
+    agent_provider: null,
+    status: 'active',
+    container_status: 'stopped',
+    last_active: now(),
+    created_at: now(),
+  });
+
+  // Authorized approver + a cached DM so ensureUserDm resolves without a
+  // platform openDM call.
+  upsertUser({ id: 'slack:admin-1', kind: 'slack', display_name: 'Admin', created_at: now() });
+  grantRole({ user_id: 'slack:admin-1', role: 'owner', agent_group_id: null, granted_by: null, granted_at: now() });
+  createMessagingGroup({
+    id: 'mg-dm-1',
+    channel_type: DM_CHANNEL,
+    platform_id: DM_PLATFORM,
+    name: 'Admin DM',
+    is_group: 0,
+    unknown_sender_policy: 'strict',
+    created_at: now(),
+  });
+  upsertUserDm({
+    user_id: 'slack:admin-1',
+    channel_type: DM_CHANNEL,
+    messaging_group_id: 'mg-dm-1',
+    resolved_at: now(),
+  });
+
+  setDeliveryAdapter(fakeAdapter);
+});
+
+afterEach(() => {
+  closeDb();
+  if (fs.existsSync(TEST_DIR)) fs.rmSync(TEST_DIR, { recursive: true, force: true });
+});
+
+describe('reject with reason', () => {
+  it('holds the row and prompts the admin instead of finalizing', async () => {
+    seedApproval('appr-1');
+    await clickRejectWithReason('appr-1');
+
+    const row = getPendingApproval('appr-1');
+    expect(row?.status).toBe('awaiting_reason');
+    expect(row?.expires_at).toBeTruthy();
+
+    // Prompt went to the admin's resolved DM, not the (empty) click platformId.
+    expect(delivered).toHaveLength(1);
+    expect(delivered[0].channelType).toBe(DM_CHANNEL);
+    expect(delivered[0].platformId).toBe(DM_PLATFORM);
+    expect((JSON.parse(delivered[0].content) as { text: string }).text).toMatch(/reason/i);
+
+    // Agent is not notified yet — the hold is still open.
+    expect(vi.mocked(writeSessionMessage)).not.toHaveBeenCalled();
+  });
+
+  it('relays the captured reason as one combined message and clears the row', async () => {
+    const { captureReasonReply } = await import('./reason-capture.js');
+    seedApproval('appr-2', 'install_packages');
+    await clickRejectWithReason('appr-2');
+
+    const consumed = await captureReasonReply(dmReply('too risky for prod'));
+
+    expect(consumed).toBe(true);
+    expect(getPendingApproval('appr-2')).toBeUndefined();
+    expect(lastRelayedText()).toBe('Your install_packages request was rejected by admin: "too risky for prod"');
+  });
+
+  it('truncates an over-long reason to 280 chars with an ellipsis', async () => {
+    const { captureReasonReply } = await import('./reason-capture.js');
+    seedApproval('appr-3');
+    await clickRejectWithReason('appr-3');
+
+    await captureReasonReply(dmReply('x'.repeat(400)));
+
+    const reason = lastRelayedText()!.match(/: "(.*)"$/)![1];
+    expect(reason).toHaveLength(280);
+    expect(reason.endsWith('…')).toBe(true);
+  });
+
+  it('finalizes a plain reject when the captured reply carries no text', async () => {
+    const { captureReasonReply } = await import('./reason-capture.js');
+    seedApproval('appr-4');
+    await clickRejectWithReason('appr-4');
+
+    const consumed = await captureReasonReply(dmReply(undefined));
+
+    expect(consumed).toBe(true);
+    expect(getPendingApproval('appr-4')).toBeUndefined();
+    expect(lastRelayedText()).toBe('Your create_agent request was rejected by admin.');
+  });
+
+  it('does not swallow a later DM once the hold was already finalized', async () => {
+    const { captureReasonReply } = await import('./reason-capture.js');
+    seedApproval('appr-5');
+    await clickRejectWithReason('appr-5');
+    // Simulate the sweep (or any other path) finalizing first.
+    deletePendingApproval('appr-5');
+
+    const consumed = await captureReasonReply(dmReply('late reason'));
+
+    expect(consumed).toBe(false);
+  });
+
+  it('ignores DMs on channels with no armed reason capture', async () => {
+    const { captureReasonReply } = await import('./reason-capture.js');
+    const consumed = await captureReasonReply({
+      channelType: DM_CHANNEL,
+      platformId: 'D-someone-else',
+      threadId: null,
+      message: { id: 'm', kind: 'chat', content: JSON.stringify({ text: 'hi' }), timestamp: now() },
+    });
+    expect(consumed).toBe(false);
+  });
+});
+
+describe('reject-with-reason host sweep', () => {
+  it('finalizes a hold whose window elapsed as a plain reject', async () => {
+    const { sweepAwaitingReasonRejects } = await import('./reason-capture.js');
+    seedApproval('appr-ghost', 'add_mcp_server');
+    markApprovalAwaitingReason('appr-ghost', new Date(Date.now() - 1000).toISOString());
+
+    await sweepAwaitingReasonRejects();
+
+    expect(getPendingApproval('appr-ghost')).toBeUndefined();
+    expect(lastRelayedText()).toBe('Your add_mcp_server request was rejected by admin.');
+  });
+
+  it('leaves a still-open hold untouched', async () => {
+    const { sweepAwaitingReasonRejects } = await import('./reason-capture.js');
+    seedApproval('appr-open');
+    markApprovalAwaitingReason('appr-open', new Date(Date.now() + 60_000).toISOString());
+
+    await sweepAwaitingReasonRejects();
+
+    expect(getPendingApproval('appr-open')?.status).toBe('awaiting_reason');
+    expect(vi.mocked(writeSessionMessage)).not.toHaveBeenCalled();
+  });
+});
+
+describe('plain reject (regression)', () => {
+  it('finalizes immediately with no reason and no DM prompt', async () => {
+    const { handleApprovalsResponse } = await import('./response-handler.js');
+    seedApproval('appr-plain', 'install_packages');
+
+    await handleApprovalsResponse({
+      questionId: 'appr-plain',
+      value: 'reject',
+      userId: 'admin-1',
+      channelType: DM_CHANNEL,
+      platformId: '',
+      threadId: null,
+    });
+
+    expect(getPendingApproval('appr-plain')).toBeUndefined();
+    expect(delivered).toHaveLength(0);
+    expect(lastRelayedText()).toBe('Your install_packages request was rejected by admin.');
+  });
+});
@@ -0,0 +1,174 @@
+/**
+ * "Reject with reason…" capture flow.
+ *
+ * When an admin clicks the third approval button, the reject is held instead of
+ * finalized: the row is parked at status='awaiting_reason' and the admin is
+ * prompted in their DM for a one-line reason. Their next DM (≤ 280 chars) is
+ * captured by a router message-interceptor and relayed to the requesting agent
+ * as one combined message — `Your <action> request was rejected by admin:
+ * "<reason>"`. A plain Reject never arms this, so an unrelated DM is never
+ * swallowed.
+ *
+ * Restart-safety: arming lives in an in-memory map (lost on restart, like the
+ * agent-naming capture it mirrors), but the hold is a durable DB row. If the
+ * admin never replies — or the host restarts mid-capture — the host sweep
+ * (sweepAwaitingReasonRejects, run each tick) finalizes a plain reject once the
+ * row's window elapses, so the requesting agent is never stranded.
+ *
+ * Reuses, not reinvents: the agent-naming prompt-then-capture pattern
+ * (in-memory map + next-DM interceptor) and the shared finalizeReject path.
+ */
+import type { InboundEvent } from '../../channels/adapter.js';
+import { getDeliveryAdapter } from '../../delivery.js';
+import {
+  deletePendingApproval,
+  getExpiredAwaitingReasonApprovals,
+  getPendingApproval,
+  getSession,
+  markApprovalAwaitingReason,
+} from '../../db/sessions.js';
+import { log } from '../../log.js';
+import { registerMessageInterceptor } from '../../router.js';
+import type { PendingApproval, Session } from '../../types.js';
+import { ensureUserDm } from '../permissions/user-dm.js';
+import { finalizeReject } from './finalize.js';
+
+/** How long an awaiting-reason hold waits for the admin's reply before the sweep finalizes a plain reject. */
+const REASON_CAPTURE_WINDOW_MS = 5 * 60 * 1000;
+/** Cap on the relayed reason — one cheap guardrail against a wall of text landing in another team's agent context. */
+const MAX_REASON_LEN = 280;
+
+const PROMPT_TEXT =
+  "Reply with a one-line reason for the rejection — I'll relay it to the agent. " +
+  'No reply within ~5 min declines it without a reason.';
+
+interface ReasonArming {
+  approvalId: string;
+  /** Namespaced id of the admin who clicked, for resolution attribution. */
+  userId: string;
+}
+
+/**
+ * Approvers waiting to type a rejection reason, keyed by their DM channel
+ * (`<channelType>:<dmPlatformId>`). A DM's platform id is unique per user, so
+ * the inbound reply matches by channel alone — no sender re-parsing needed, and
+ * a group message can never collide with an armed DM. Cleared on receipt,
+ * staleness, or restart.
+ */
+const awaitingReason = new Map<string, ReasonArming>();
+
+function dmKey(channelType: string, platformId: string): string {
+  return `${channelType}:${platformId}`;
+}
+
+function clampReason(raw: string): string {
+  const trimmed = raw.trim();
+  if (trimmed.length <= MAX_REASON_LEN) return trimmed;
+  return trimmed.slice(0, MAX_REASON_LEN - 1) + '…';
+}
+
+function extractText(event: InboundEvent): string {
+  try {
+    const parsed = JSON.parse(event.message.content) as Record<string, unknown>;
+    return typeof parsed.text === 'string' ? parsed.text : '';
+  } catch {
+    return '';
+  }
+}
+
+/**
+ * Begin the reject-with-reason hold for an approval the admin chose not to
+ * finalize outright. Prompts the admin's DM, then parks the row and arms
+ * capture. If we can't reach the admin (no DM, no adapter, delivery throws) we
+ * finalize a plain reject immediately rather than strand the requesting agent.
+ */
+export async function armReasonCapture(approval: PendingApproval, session: Session, userId: string): Promise<void> {
+  const dm = userId ? await ensureUserDm(userId) : null;
+  const adapter = getDeliveryAdapter();
+  if (!dm || !adapter) {
+    log.warn('reject-with-reason: cannot reach approver, finalizing plain reject', {
+      approvalId: approval.approval_id,
+      userId,
+      hasDm: Boolean(dm),
+      hasAdapter: Boolean(adapter),
+    });
+    await finalizeReject(approval, session, userId);
+    return;
+  }
+
+  try {
+    await adapter.deliver(dm.channel_type, dm.platform_id, null, 'chat-sdk', JSON.stringify({ text: PROMPT_TEXT }));
+  } catch (err) {
+    log.error('reject-with-reason: reason prompt delivery failed, finalizing plain reject', {
+      approvalId: approval.approval_id,
+      err,
+    });
+    await finalizeReject(approval, session, userId);
+    return;
+  }
+
+  // Prompt is out — now hold the row and arm capture. Order matters: a reply
+  // can't arrive before the prompt is read, so there's no lost-message window.
+  const expiresAt = new Date(Date.now() + REASON_CAPTURE_WINDOW_MS).toISOString();
+  markApprovalAwaitingReason(approval.approval_id, expiresAt);
+  awaitingReason.set(dmKey(dm.channel_type, dm.platform_id), { approvalId: approval.approval_id, userId });
+  log.info('reject-with-reason: awaiting reason reply', { approvalId: approval.approval_id, userId });
+}
+
+/**
+ * Router message-interceptor: capture the next DM from an admin who armed a
+ * reason. Returns true (consume the message) when this DM is an armed reason
+ * channel and still holds a live row; false otherwise so normal routing runs.
+ *
+ * Exported for tests; registered as the interceptor below.
+ */
+export async function captureReasonReply(event: InboundEvent): Promise<boolean> {
+  const arming = awaitingReason.get(dmKey(event.channelType, event.platformId));
+  if (!arming) return false;
+
+  // This DM is an armed reason channel — disarm regardless of outcome.
+  awaitingReason.delete(dmKey(event.channelType, event.platformId));
+
+  const approval = getPendingApproval(arming.approvalId);
+  if (!approval || approval.status !== 'awaiting_reason') {
+    // Already finalized (e.g. ghosted by the sweep). The reply is no longer a
+    // reason — let it route normally instead of swallowing it.
+    return false;
+  }
+
+  const session = approval.session_id ? getSession(approval.session_id) : null;
+  if (!session) {
+    deletePendingApproval(approval.approval_id);
+    return true;
+  }
+
+  const reason = clampReason(extractText(event));
+  await finalizeReject(approval, session, arming.userId, reason || undefined);
+  log.info('reject-with-reason: reason captured and relayed', {
+    approvalId: approval.approval_id,
+    hasReason: reason.length > 0,
+  });
+  return true;
+}
+
+registerMessageInterceptor(captureReasonReply);
+
+/**
+ * Host-sweep finalizer: any reject-with-reason hold whose window elapsed (admin
+ * ghosted, or the host restarted mid-capture and lost the in-memory arming) is
+ * finalized as a plain reject. Restart-safe — the hold is a durable row, so the
+ * requesting agent always gets its decision. Called once per sweep tick.
+ */
+export async function sweepAwaitingReasonRejects(): Promise<void> {
+  const rows = getExpiredAwaitingReasonApprovals(new Date().toISOString());
+  for (const approval of rows) {
+    const session = approval.session_id ? getSession(approval.session_id) : null;
+    if (!session) {
+      deletePendingApproval(approval.approval_id);
+      continue;
+    }
+    // Plain reject, unknown resolver — the admin opted in but never typed.
+    await finalizeReject(approval, session, '');
+    log.info('reject-with-reason: window elapsed, finalized as plain reject', { approvalId: approval.approval_id });
+  }
+}
@@ -5,7 +5,10 @@
 *   1. Module-initiated actions — the module called `requestApproval()` with
 *      some free-form `action` string and registered a handler via
 *      `registerApprovalHandler(action, handler)`. On approve, we look up the
- *      handler and call it; on reject, we notify the agent and move on.
+ *      handler and call it; on plain reject we relay a decline to the agent; on
+ *      "Reject with reason…" we hold the row and capture the admin's next DM as
+ *      a one-line reason (see reason-capture.ts). Reject finalization is shared
+ *      via finalizeReject.
 *   2. OneCLI credential approvals (`action = 'onecli_credential'`). Resolved
 *      via an in-memory Promise — see onecli-approvals.ts.
 *
@@ -19,8 +22,10 @@ import { log } from '../../log.js';
 import { writeSessionMessage } from '../../session-manager.js';
 import type { PendingApproval } from '../../types.js';
 import { hasAdminPrivilege, isGlobalAdmin, isOwner } from '../permissions/db/user-roles.js';
+import { finalizeReject } from './finalize.js';
 import { ONECLI_ACTION, resolveOneCLIApproval } from './onecli-approvals.js';
-import { getApprovalHandler, notifyApprovalResolved } from './primitive.js';
+import { getApprovalHandler, notifyApprovalResolved, REJECT_WITH_REASON_VALUE } from './primitive.js';
+import { armReasonCapture } from './reason-capture.js';

 export async function handleApprovalsResponse(payload: ResponsePayload): Promise<boolean> {
  const approval = getPendingApproval(payload.questionId);
@@ -65,6 +70,21 @@ async function handleRegisteredApproval(
    return;
  }

+  // "Reject with reason…" — hold the row and capture the admin's next DM
+  // instead of finalizing now. The agent is notified exactly once: after the
+  // reason arrives, or after the sweep's timeout if the admin ghosts.
+  if (selectedOption === REJECT_WITH_REASON_VALUE) {
+    await armReasonCapture(approval, session, userId);
+    return;
+  }
+
+  // Plain Reject (or any other non-approve value) — instant fast path.
+  if (selectedOption !== 'approve') {
+    await finalizeReject(approval, session, userId);
+    return;
+  }
+
+  // Approved — dispatch to the module that registered for this action.
  const notify = (text: string): void => {
    writeSessionMessage(session.agent_group_id, session.id, {
      id: `appr-note-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
@@ -77,16 +97,6 @@ async function handleRegisteredApproval(
    });
  };

-  if (selectedOption !== 'approve') {
-    notify(`Your ${approval.action} request was rejected by admin.`);
-    log.info('Approval rejected', { approvalId: approval.approval_id, action: approval.action, userId });
-    deletePendingApproval(approval.approval_id);
-    await notifyApprovalResolved({ approval, session, outcome: 'reject', userId });
-    await wakeContainer(session);
-    return;
-  }
-
-  // Approved — dispatch to the module that registered for this action.
  const handler = getApprovalHandler(approval.action);
  if (!handler) {
    log.warn('No approval handler registered — row dropped', {
@@ -22,7 +22,7 @@ import {
  routeInbound,
  setAccessGate,
  setChannelRequestGate,
-  setMessageInterceptor,
+  registerMessageInterceptor,
  setSenderResolver,
  setSenderScopeGate,
  type AccessGateResult,
@@ -521,7 +521,7 @@ registerResponseHandler(handleChannelApprovalResponse);
 // Captures the next DM from an approver who clicked "Create new agent",
 // creates the agent immediately, wires the channel, and replays.

-setMessageInterceptor(async (event: InboundEvent): Promise<boolean> => {
+registerMessageInterceptor(async (event: InboundEvent): Promise<boolean> => {
  const userId = extractAndUpsertUser(event);
  if (!userId) return false;

@@ -110,16 +110,20 @@ export function setSenderScopeGate(fn: SenderScopeGateFn): void {

 /**
 * Message-interceptor hook. Runs at the very top of routeInbound, before
- * messaging-group resolution. When the interceptor returns true the message
- * is consumed and routing stops. Used by the permissions module to capture
- * free-text replies during multi-step approval flows (e.g. agent naming).
+ * messaging-group resolution. When an interceptor returns true the message is
+ * consumed and routing stops. Multiple interceptors may register; they run in
+ * registration order and the first to claim the message (return true) wins.
+ *
+ * Used by modules to capture free-text DM replies during multi-step approval
+ * flows — the permissions module (agent naming during channel registration)
+ * and the approvals module (reject-with-reason capture).
 */
 export type MessageInterceptorFn = (event: InboundEvent) => Promise<boolean>;

-let messageInterceptor: MessageInterceptorFn | null = null;
+const messageInterceptors: MessageInterceptorFn[] = [];

-export function setMessageInterceptor(fn: MessageInterceptorFn): void {
-  messageInterceptor = fn;
+export function registerMessageInterceptor(fn: MessageInterceptorFn): void {
+  messageInterceptors.push(fn);
 }

 /**
@@ -156,9 +160,13 @@ function safeParseContent(raw: string): { text?: string; sender?: string; sender
 * Creates messaging group + session if they don't exist yet.
 */
 export async function routeInbound(event: InboundEvent): Promise<void> {
-  // Pre-route interceptor — lets modules consume messages before any routing
-  // (e.g. free-text replies during multi-step approval flows).
-  if (messageInterceptor && (await messageInterceptor(event))) return;
+  // Pre-route interceptors — let modules consume messages before any routing
+  // (e.g. free-text DM replies during multi-step approval flows). They run in
+  // registration order; the first to claim the message stops routing. The
+  // sequential await is intentional — first-to-claim is order-dependent.
+  for (const intercept of messageInterceptors) {
+    if (await intercept(event)) return;
+  }

  // 0. Apply the adapter's thread policy. Non-threaded adapters (Telegram,
  //    WhatsApp, iMessage, email) collapse threads to the channel. Resolved
@@ -0,0 +1,103 @@
+/**
+ * Security regression for the channel-inbound attachment path (#2828 sibling).
+ *
+ * `extractAttachmentFiles` (via `writeSessionMessage`) hardens the per-message
+ * inbox subdir against pre-placed symlinks, but NOT the `inbox` root itself.
+ * A compromised container can write inside its own session dir, so it can
+ * replace `inbox` with a symlink pointing outside the session sandbox. The
+ * existing guard then:
+ *   - skips the lstat branch (it only lstats `inbox/<msgId>`, not `inbox`),
+ *   - mkdirs `inbox/<msgId>` *through* the symlink,
+ *   - passes the containment check, because it compares against
+ *     `realpathSync(inboxRoot)` which has already followed the symlink, and
+ *   - writes a brand-new file (the `wx` flag only blocks an existing dst).
+ *
+ * Result: the host writes attacker-influenced bytes outside the session root —
+ * the same class of bug fixed for the A2A path in forwardAttachedFiles (#2828).
+ *
+ * This test asserts the SECURE behaviour (nothing written outside). It FAILS
+ * against the current code, demonstrating the gap.
+ */
+import fs from 'fs';
+import path from 'path';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+vi.mock('./config.js', async () => {
+  const actual = await vi.importActual<typeof import('./config.js')>('./config.js');
+  return { ...actual, DATA_DIR: '/tmp/nanoclaw-test-saveatt-gap' };
+});
+
+import { initTestDb, closeDb, runMigrations, createAgentGroup } from './db/index.js';
+import { createSession } from './db/sessions.js';
+import { initSessionFolder, sessionDir, writeSessionMessage } from './session-manager.js';
+import type { Session } from './types.js';
+
+const TEST_DIR = '/tmp/nanoclaw-test-saveatt-gap';
+const AG = 'ag-saveatt';
+const SESS = 'sess-saveatt';
+
+function now(): string {
+  return new Date().toISOString();
+}
+
+beforeEach(() => {
+  if (fs.existsSync(TEST_DIR)) fs.rmSync(TEST_DIR, { recursive: true });
+  fs.mkdirSync(TEST_DIR, { recursive: true });
+
+  const db = initTestDb();
+  runMigrations(db);
+
+  createAgentGroup({ id: AG, name: 'SaveAtt', folder: 'saveatt', agent_provider: null, created_at: now() });
+  const sess: Session = {
+    id: SESS,
+    agent_group_id: AG,
+    messaging_group_id: null,
+    thread_id: null,
+    agent_provider: null,
+    status: 'active',
+    container_status: 'stopped',
+    last_active: null,
+    created_at: now(),
+  };
+  createSession(sess);
+  initSessionFolder(AG, SESS);
+});
+
+afterEach(() => {
+  closeDb();
+  if (fs.existsSync(TEST_DIR)) fs.rmSync(TEST_DIR, { recursive: true });
+});
+
+describe('extractAttachmentFiles — inbox-root symlink containment (#2828 sibling)', () => {
+  it('does not write an attachment outside the session root via a symlinked inbox root', () => {
+    // Attacker-controlled location outside the session sandbox.
+    const canaryDir = path.join(TEST_DIR, 'canary-outside');
+    fs.mkdirSync(canaryDir, { recursive: true });
+
+    // Container pre-places its whole `inbox` as a symlink pointing outside.
+    const inboxRoot = path.join(sessionDir(AG, SESS), 'inbox');
+    fs.rmSync(inboxRoot, { recursive: true, force: true });
+    fs.symlinkSync(canaryDir, inboxRoot);
+
+    const content = JSON.stringify({
+      text: 'see attached',
+      attachments: [{ name: 'pwn.txt', data: Buffer.from('attacker-bytes').toString('base64') }],
+    });
+
+    writeSessionMessage(AG, SESS, {
+      id: 'evil-inbox-root',
+      kind: 'chat',
+      timestamp: now(),
+      platformId: 'whatsapp:123',
+      channelType: 'whatsapp',
+      threadId: null,
+      content,
+    });
+
+    // SECURE expectation: nothing was written through the symlink to the
+    // attacker-controlled canary location.
+    const escaped = path.join(canaryDir, 'evil-inbox-root', 'pwn.txt');
+    expect(fs.existsSync(escaped)).toBe(false);
+    expect(fs.readdirSync(canaryDir)).toHaveLength(0);
+  });
+});
@@ -18,6 +18,7 @@ import { deriveAttachmentName } from './attachment-naming.js';
 import { isSafeAttachmentName } from './attachment-safety.js';
 import type { OutboundFile } from './channels/adapter.js';
 import { DATA_DIR } from './config.js';
+import { ensureContainedInboxDir, isPathInside } from './inbox-safety.js';
 import { getMessagingGroup } from './db/messaging-groups.js';
 import {
  createSession,
@@ -38,11 +39,6 @@ import {
 import { log } from './log.js';
 import type { Session } from './types.js';

-function isPathInside(parent: string, child: string): boolean {
-  const relative = path.relative(parent, child);
-  return relative === '' || (!relative.startsWith('..') && !path.isAbsolute(relative));
-}
-
 /** Root directory for all session data. */
 export function sessionsBaseDir(): string {
  return path.join(DATA_DIR, 'v2-sessions');
@@ -288,6 +284,14 @@ function extractAttachmentFiles(
    return contentStr;
  }

+  const inboxRoot = path.join(sessionDir(agentGroupId, sessionId), 'inbox');
+  // Resolved lazily on the first attachment that actually carries bytes, so a
+  // message whose attachments have no inline `data` never creates an inbox dir.
+  // ensureContainedInboxDir refuses a pre-placed symlink at the inbox root or
+  // the per-message subdir before any write lands outside the sandbox (#2828).
+  let inboxDir: string | null = null;
+  let inboxResolved = false;
+
  let changed = false;
  for (const att of attachments) {
    if (typeof att.data !== 'string') continue;
@@ -302,32 +306,12 @@ function extractAttachmentFiles(
      });
    }

-    const inboxDir = path.join(sessionDir(agentGroupId, sessionId), 'inbox', messageId);
-
-    // Refuse to mkdir through a symlink that the container may have pre placed
-    // at inboxDir. With recursive:true, mkdirSync would silently no op on a
-    // pre existing symlink and the subsequent writeFileSync would follow it.
-    if (fs.existsSync(inboxDir)) {
-      const stat = fs.lstatSync(inboxDir);
-      if (stat.isSymbolicLink() || !stat.isDirectory()) {
-        log.warn('Rejecting unsafe inbox directory', { messageId, inboxDir });
-        continue;
-      }
-    }
-    fs.mkdirSync(inboxDir, { recursive: true });
-
-    let realInboxDir: string;
-    try {
-      realInboxDir = fs.realpathSync(inboxDir);
-    } catch (err) {
-      log.warn('Failed to resolve inbox directory', { messageId, err });
-      continue;
-    }
-    const inboxRoot = path.join(sessionDir(agentGroupId, sessionId), 'inbox');
-    if (!isPathInside(fs.realpathSync(inboxRoot), realInboxDir)) {
-      log.warn('Inbox directory escaped session inbox root', { messageId, inboxDir });
-      continue;
+    if (!inboxResolved) {
+      inboxDir = ensureContainedInboxDir(inboxRoot, messageId, { messageId });
+      inboxResolved = true;
    }
+    // Unsafe inbox (symlink / escape) — no attachment can be written safely.
+    if (!inboxDir) break;

    const filePath = path.join(inboxDir, filename);
    try {
@@ -0,0 +1,83 @@
+import fs from 'fs';
+import path from 'path';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+const TEST_ROOT = '/tmp/nanoclaw-create-agent-test';
+const GROUPS_DIR = path.join(TEST_ROOT, 'groups');
+const DATA_DIR = path.join(TEST_ROOT, 'data');
+const TEMPLATES_DIR = path.join(TEST_ROOT, 'templates');
+
+vi.mock('../config.js', async (importOriginal) => ({
+  ...(await importOriginal<typeof import('../config.js')>()),
+  GROUPS_DIR: '/tmp/nanoclaw-create-agent-test/groups',
+  DATA_DIR: '/tmp/nanoclaw-create-agent-test/data',
+  TEMPLATES_DIR: '/tmp/nanoclaw-create-agent-test/templates',
+}));
+
+vi.mock('../log.js', () => ({
+  log: { debug: vi.fn(), info: vi.fn(), warn: vi.fn(), error: vi.fn(), fatal: vi.fn() },
+}));
+
+import { closeDb, initTestDb, runMigrations } from '../db/index.js';
+import { getContainerConfig } from '../db/container-configs.js';
+import { PERSONA_PREPEND_FILE } from '../group-persona.js';
+import { createAgentFromTemplate } from './create-agent.js';
+
+function writeTemplate(): void {
+  const t = path.join(TEMPLATES_DIR, 'sales', 'sdr');
+  fs.mkdirSync(path.join(t, 'context', 'additional_context'), { recursive: true });
+  fs.writeFileSync(path.join(t, 'context', 'instructions.md'), 'You are an SDR agent.\n');
+  fs.writeFileSync(path.join(t, 'context', 'playbook.md'), '# Playbook\n');
+  fs.writeFileSync(path.join(t, 'context', 'additional_context', 'faq.md'), '# FAQ\n');
+  fs.writeFileSync(
+    path.join(t, '.mcp.json'),
+    JSON.stringify({ mcpServers: { hubspot: { command: 'npx', args: ['-y', '@hubspot/mcp-server'] } } }),
+  );
+  const skillDir = path.join(t, 'skills', 'widget');
+  fs.mkdirSync(skillDir, { recursive: true });
+  fs.writeFileSync(path.join(skillDir, 'SKILL.md'), '---\nname: widget\n---\n');
+}
+
+beforeEach(() => {
+  fs.rmSync(TEST_ROOT, { recursive: true, force: true });
+  fs.mkdirSync(TEST_ROOT, { recursive: true });
+  runMigrations(initTestDb());
+  writeTemplate();
+});
+
+afterEach(() => {
+  closeDb();
+  fs.rmSync(TEST_ROOT, { recursive: true, force: true });
+});
+
+describe('createAgentFromTemplate', () => {
+  it('writes the persona prepend verbatim — no injected context refs, no .seed.md', () => {
+    const g = createAgentFromTemplate('sales/sdr', { name: 'SDR Test' });
+
+    const groupDir = path.join(GROUPS_DIR, g.folder);
+    const prepend = fs.readFileSync(path.join(groupDir, PERSONA_PREPEND_FILE), 'utf-8');
+    expect(prepend).toBe('You are an SDR agent.\n');
+    expect(fs.existsSync(path.join(groupDir, '.seed.md'))).toBe(false);
+  });
+
+  it('copies template skills into the group-private Claude-plane skills dir', () => {
+    const g = createAgentFromTemplate('sales/sdr', { name: 'SDR Skills' });
+
+    const skill = path.join(DATA_DIR, 'v2-sessions', g.id, '.claude-shared', 'skills', 'widget', 'SKILL.md');
+    expect(fs.existsSync(skill)).toBe(true);
+  });
+
+  it('writes MCP servers to the container config and context extras at their template-relative paths', () => {
+    const g = createAgentFromTemplate('sales/sdr', { name: 'SDR Mcp' });
+
+    const cfg = getContainerConfig(g.id);
+    expect(cfg).toBeTruthy();
+    expect(JSON.parse(cfg!.mcp_servers)).toHaveProperty('hubspot');
+    // Extras land relative to the group root, exactly as they sit relative to
+    // instructions.md in the template — no context/ prefix in between.
+    const groupDir = path.join(GROUPS_DIR, g.folder);
+    expect(fs.existsSync(path.join(groupDir, 'playbook.md'))).toBe(true);
+    expect(fs.existsSync(path.join(groupDir, 'additional_context', 'faq.md'))).toBe(true);
+    expect(fs.existsSync(path.join(groupDir, 'context'))).toBe(false);
+  });
+});
@@ -0,0 +1,78 @@
+import { randomUUID } from 'crypto';
+import fs from 'fs';
+import path from 'path';
+
+import { DATA_DIR, GROUPS_DIR } from '../config.js';
+import { createAgentGroup } from '../db/agent-groups.js';
+import { ensureContainerConfig, updateContainerConfigJson } from '../db/container-configs.js';
+import { assertValidGroupFolder, resolveGroupFolderPath } from '../group-folder.js';
+import { PERSONA_PREPEND_FILE } from '../group-persona.js';
+import { normalizeName } from '../modules/agent-to-agent/db/agent-destinations.js';
+import type { AgentGroup } from '../types.js';
+import { resolveLocalTemplate } from './local-dir.js';
+import { parseTemplate } from './parse.js';
+
+export interface CreateAgentOptions {
+  name?: string;
+}
+
+/**
+ * Stamp a self-contained agent group from a LOCAL template ref under
+ * TEMPLATES_DIR. The template carries MCP servers, instructions, optional
+ * context extras, and optional skills — nothing else (no policy, no packages,
+ * no provider).
+ *
+ * The template persona is written to the provider-neutral `instructions.prepend.md`
+ * (see src/group-persona.ts). Each provider's project-doc composer inlines it at
+ * the TOP of the doc it generates every spawn, so the persona is system-prompt
+ * tier regardless of which provider the group ends up running. Because the file
+ * is provider-agnostic, placement needs no provider knowledge at stamp time (the
+ * provider is DB-resolved later, at first spawn).
+ *
+ * Returns the created group; the caller wires it to a channel as usual.
+ */
+export function createAgentFromTemplate(ref: string, opts?: CreateAgentOptions): AgentGroup {
+  const dir = resolveLocalTemplate(ref);
+  const tpl = parseTemplate(dir);
+
+  const id = randomUUID();
+  const name = opts?.name ?? path.basename(dir);
+  let folder = normalizeName(name);
+  assertValidGroupFolder(folder);
+  if (fs.existsSync(resolveGroupFolderPath(folder))) folder = `${folder}-${randomUUID().slice(0, 8)}`;
+
+  const group: AgentGroup = { id, name, folder, agent_provider: null, created_at: new Date().toISOString() };
+  createAgentGroup(group);
+  ensureContainerConfig(id);
+
+  // group-init.ts owns the mkdir at first spawn, but it isn't called here — so we
+  // create the dir ourselves to land instructions.prepend.md + context/.
+  const groupDir = path.resolve(GROUPS_DIR, folder);
+  fs.mkdirSync(groupDir, { recursive: true });
+
+  // Persona → provider-neutral prepend, inlined at the top of the group's
+  // CLAUDE.md/AGENTS.md every spawn (system-prompt tier on any provider).
+  fs.writeFileSync(path.join(groupDir, PERSONA_PREPEND_FILE), tpl.instructions + '\n');
+
+  // Context extras keep their template-relative layout, placed next to the doc
+  // the persona is inlined into — so a reference written in instructions.md
+  // (e.g. `additional_context/faq.md`) resolves unchanged in the agent's
+  // workspace. Nothing is injected into the persona; referencing each file from
+  // instructions.md is the template author's job (docs/templates.md).
+  for (const { name: file, content } of tpl.contextExtras) {
+    const dest = path.join(groupDir, file);
+    fs.mkdirSync(path.dirname(dest), { recursive: true });
+    fs.writeFileSync(dest, content);
+  }
+
+  updateContainerConfigJson(id, 'mcp_servers', tpl.mcpServers);
+
+  // Per-group skills overlay — keyed by group id, never shared. cpSync creates
+  // intermediate dirs, so .claude-shared/skills need not exist yet.
+  const skillsDir = path.join(DATA_DIR, 'v2-sessions', id, '.claude-shared', 'skills');
+  for (const { name: skill, srcDir } of tpl.skills) {
+    fs.cpSync(srcDir, path.join(skillsDir, skill), { recursive: true });
+  }
+
+  return group;
+}
@@ -0,0 +1,55 @@
+import fs from 'fs';
+import os from 'os';
+import path from 'path';
+
+import { afterEach, beforeEach, describe, expect, it } from 'vitest';
+
+import { resolveLocalTemplate } from './local-dir.js';
+
+let base: string;
+
+beforeEach(() => {
+  base = fs.mkdtempSync(path.join(os.tmpdir(), 'tpl-local-'));
+  fs.mkdirSync(path.join(base, 'sales', 'sdr'), { recursive: true });
+  fs.writeFileSync(path.join(base, 'afile.md'), 'not a directory');
+});
+afterEach(() => fs.rmSync(base, { recursive: true, force: true }));
+
+describe('resolveLocalTemplate', () => {
+  it('resolves a valid multi-segment relative ref under the base', () => {
+    expect(resolveLocalTemplate('sales/sdr', base)).toBe(path.join(base, 'sales', 'sdr'));
+  });
+
+  it('rejects a ref that escapes the base via ../', () => {
+    expect(() => resolveLocalTemplate('../escape', base)).toThrow(/escapes/);
+  });
+
+  it('rejects a multi-segment escape like sales/../../etc', () => {
+    expect(() => resolveLocalTemplate('sales/../../etc', base)).toThrow(/escapes/);
+  });
+
+  it('rejects an absolute ref', () => {
+    expect(() => resolveLocalTemplate('/etc', base)).toThrow(/relative/);
+  });
+
+  it('rejects a ~-prefixed ref', () => {
+    expect(() => resolveLocalTemplate('~/x', base)).toThrow(/relative/);
+  });
+
+  it('rejects empty and whitespace-only refs', () => {
+    expect(() => resolveLocalTemplate('', base)).toThrow(/Invalid/);
+    expect(() => resolveLocalTemplate('   ', base)).toThrow(/Invalid/);
+  });
+
+  it('rejects an untrimmed ref', () => {
+    expect(() => resolveLocalTemplate(' sales/sdr', base)).toThrow(/Invalid/);
+  });
+
+  it('throws when the ref does not exist', () => {
+    expect(() => resolveLocalTemplate('nope', base)).toThrow(/not found/i);
+  });
+
+  it('throws when the ref is a file, not a directory', () => {
+    expect(() => resolveLocalTemplate('afile.md', base)).toThrow(/not found/i);
+  });
+});
@@ -0,0 +1,33 @@
+import fs from 'fs';
+import path from 'path';
+
+import { TEMPLATES_DIR } from '../config.js';
+
+/**
+ * Resolve a LOCAL template ref to an absolute directory under `base`
+ * (TEMPLATES_DIR by default). Lexical containment only — no realpathSync, no
+ * symlink resolution (out of threat model). Mirrors ensureWithinBase() in
+ * group-folder.ts. Refs are legitimately multi-segment (e.g. "sales/sdr"), so
+ * this does NOT reuse isValidGroupFolder (which rejects "/").
+ *
+ * Rejects: empty / untrimmed refs, absolute paths, a leading "~", and any ref
+ * that escapes `base` after resolution. Throws if the resolved path is missing
+ * or not a directory.
+ */
+export function resolveLocalTemplate(ref: string, base: string = TEMPLATES_DIR): string {
+  if (!ref || ref !== ref.trim()) {
+    throw new Error(`Invalid template ref: "${ref}"`);
+  }
+  if (path.isAbsolute(ref) || ref.startsWith('~')) {
+    throw new Error(`Template ref must be relative to the templates directory: "${ref}"`);
+  }
+  const candidate = path.resolve(base, ref);
+  const rel = path.relative(base, candidate);
+  if (rel.startsWith('..') || path.isAbsolute(rel)) {
+    throw new Error(`Template ref escapes the templates directory: "${ref}"`);
+  }
+  if (!fs.existsSync(candidate) || !fs.statSync(candidate).isDirectory()) {
+    throw new Error(`Template not found: "${ref}" (looked in ${base})`);
+  }
+  return candidate;
+}
@@ -0,0 +1,56 @@
+import fs from 'fs';
+import os from 'os';
+import path from 'path';
+import { afterEach, beforeEach, describe, expect, it } from 'vitest';
+
+import { parseTemplate } from './parse.js';
+
+let dir: string;
+
+beforeEach(() => {
+  dir = fs.mkdtempSync(path.join(os.tmpdir(), 'tpl-parse-'));
+});
+afterEach(() => {
+  fs.rmSync(dir, { recursive: true, force: true });
+});
+
+function write(rel: string, content: string): void {
+  const full = path.join(dir, rel);
+  fs.mkdirSync(path.dirname(full), { recursive: true });
+  fs.writeFileSync(full, content);
+}
+
+describe('parseTemplate', () => {
+  it('parses mcpServers, instructions, context extras, and skills', () => {
+    write('.mcp.json', JSON.stringify({ mcpServers: { fs: { command: 'mcp-fs', args: ['/data'] } } }));
+    write('context/instructions.md', 'Be helpful.\n\n');
+    write('context/playbook.md', '# Playbook');
+    write('context/additional_context/faq.md', '# FAQ');
+    write('skills/research/SKILL.md', 'do research');
+    fs.writeFileSync(path.join(dir, 'context', 'notes.txt'), 'ignored'); // non-.md is ignored
+
+    const tpl = parseTemplate(dir);
+
+    expect(tpl.mcpServers).toEqual({ fs: { command: 'mcp-fs', args: ['/data'] } });
+    expect(tpl.instructions).toBe('Be helpful.'); // trimEnd, instructions.md excluded from extras
+    // Nested extras keep their context/-relative path as the name.
+    expect(tpl.contextExtras.map((c) => c.name).sort()).toEqual(['additional_context/faq.md', 'playbook.md']);
+    expect(tpl.skills.map((s) => s.name)).toEqual(['research']);
+  });
+
+  it('defaults the optionals when only instructions.md is present', () => {
+    write('context/instructions.md', 'Only instructions.');
+    const tpl = parseTemplate(dir);
+    expect(tpl.mcpServers).toEqual({});
+    expect(tpl.contextExtras).toEqual([]);
+    expect(tpl.skills).toEqual([]);
+  });
+
+  it('throws when context/instructions.md is missing', () => {
+    expect(() => parseTemplate(dir)).toThrow(/instructions\.md/);
+  });
+
+  it('throws when the folder does not exist', () => {
+    expect(() => parseTemplate(path.join(dir, 'nope'))).toThrow(/not found/i);
+  });
+});
@@ -0,0 +1,64 @@
+import fs from 'fs';
+import path from 'path';
+
+/** A parsed template folder. Pure data — no DB, no side effects. */
+export interface Template {
+  mcpServers: Record<string, unknown>; // .mcp.json .mcpServers — name -> launch config
+  instructions: string; // context/instructions.md (required)
+  contextExtras: { name: string; content: string }[]; // context/**/*.md except instructions.md; name relative to context/
+  skills: { name: string; srcDir: string }[]; // skills/<name>/ real folders
+}
+
+function readJson(file: string): unknown {
+  return fs.existsSync(file) ? JSON.parse(fs.readFileSync(file, 'utf-8')) : undefined;
+}
+
+function asRecord(value: unknown): Record<string, unknown> {
+  return value && typeof value === 'object' && !Array.isArray(value) ? (value as Record<string, unknown>) : {};
+}
+
+/**
+ * Read and lightly validate a template folder into a typed object. Throws only
+ * if the folder is missing or `context/instructions.md` (the one required file)
+ * is absent. `unknown`-in / parsed-out at the .mcp.json boundary.
+ */
+export function parseTemplate(dir: string): Template {
+  if (!fs.existsSync(dir)) throw new Error(`Template folder not found: ${dir}`);
+
+  const mcpServers = asRecord(asRecord(readJson(path.join(dir, '.mcp.json'))).mcpServers);
+
+  const instructionsFile = path.join(dir, 'context', 'instructions.md');
+  if (!fs.existsSync(instructionsFile)) {
+    throw new Error(`Template missing required context/instructions.md: ${dir}`);
+  }
+  const instructions = fs.readFileSync(instructionsFile, 'utf-8').trimEnd();
+
+  return {
+    mcpServers,
+    instructions,
+    contextExtras: readContextExtras(path.join(dir, 'context')),
+    skills: readSkills(path.join(dir, 'skills')),
+  };
+}
+
+/**
+ * Every context/**\/*.md except the top-level instructions.md, recursively.
+ * `name` keeps the path relative to context/ so stamping can preserve the
+ * layout — a reference like `additional_context/faq.md` written in
+ * instructions.md resolves unchanged in the agent's workspace.
+ */
+function readContextExtras(contextDir: string): { name: string; content: string }[] {
+  if (!fs.existsSync(contextDir)) return [];
+  return (fs.readdirSync(contextDir, { recursive: true }) as string[])
+    .filter((f) => f.endsWith('.md') && f !== 'instructions.md' && fs.statSync(path.join(contextDir, f)).isFile())
+    .map((name) => ({ name, content: fs.readFileSync(path.join(contextDir, name), 'utf-8') }));
+}
+
+/** Each immediate subdirectory of skills/ is a packaged skill. */
+function readSkills(skillsDir: string): { name: string; srcDir: string }[] {
+  if (!fs.existsSync(skillsDir)) return [];
+  return fs
+    .readdirSync(skillsDir)
+    .map((name) => ({ name, srcDir: path.join(skillsDir, name) }))
+    .filter(({ srcDir }) => fs.statSync(srcDir).isDirectory());
+}
@@ -200,8 +200,13 @@ export interface PendingApproval {
  channel_type: string | null;
  platform_id: string | null;
  platform_message_id: string | null;
+  /**
+   * For OneCLI credential rows, the gateway's request TTL. For a module
+   * approval held by "Reject with reason…", the deadline after which the
+   * host sweep finalizes a plain reject (set by markApprovalAwaitingReason).
+   */
  expires_at: string | null;
-  status: 'pending' | 'approved' | 'rejected' | 'expired';
+  status: 'pending' | 'approved' | 'rejected' | 'expired' | 'awaiting_reason';
  title: string;
  options_json: string;
  /** When set, only this exact user may resolve the approval. */
@@ -0,0 +1,48 @@
+# Templates
+
+Local agent-template library for this NanoClaw install. **This folder ships
+empty.** Anything you drop here is a template you can stamp into an agent:
+
+```bash
+ncl groups create --template <relative-ref> --name "My Agent"
+```
+
+`<relative-ref>` is a path *relative to this folder* (e.g. `sales/sdr`). Refs
+must stay inside this directory — absolute paths, `~`, and `../` escapes are
+rejected. Override the location with `NANOCLAW_TEMPLATES_DIR=/another/local/path`
+(a local path only — never a URL).
+
+The setup wizard's **Template setup → NanoClaw template library** option clones
+the public registry and copies your chosen template *into this folder*, after
+which it stamps from the local copy. **Local templates** lists whatever is here.
+
+## Anatomy of a template
+
+Only `context/instructions.md` is required; it both supplies the agent's
+standing brief and marks the folder as a template.
+
+```
+<template>/
+├── context/
+│   ├── instructions.md        # REQUIRED: the agent's standing persona, prepended to its
+│   │                          #           CLAUDE.md/AGENTS.md every spawn
+│   └── additional_context/    # optional: extra .md files
+│       └── *.md
+├── .mcp.json             # optional: { "mcpServers": { ... } } — command + args, NO secrets
+├── skills/<name>/        # optional: one folder per skill (SKILL.md + references/), copied whole
+└── README.md             # recommended: per-template docs
+```
+
+Notes:
+- **Extra context is copied preserving its layout relative to `instructions.md`**
+  (`context/additional_context/faq.md` → `additional_context/faq.md` in the
+  agent's workspace). Nothing is referenced automatically — `instructions.md`
+  must point to each file (e.g. "Pricing rules live in
+  `additional_context/pricing.md`").
+- **No provider, no model, no packages.** A template is instructions + MCP
+  servers + skills. The agent's runtime/provider is chosen separately
+  (`ncl groups config update --provider …` or during setup).
+- **No secrets.** `.mcp.json` carries launch config only; credentials are
+  injected by the credentials proxy at request time. If an MCP server refuses
+  to boot without an env var, use a placeholder value — never a real key.
+- Skills are copied into the agent's own per-group overlay, never shared.
Author	SHA1	Message	Date
github-actions[bot]	aecad864e6	chore: bump version to 2.1.24	2026-07-02 11:57:00 +00:00
gavrielc	c87f2e55dc	Merge pull request #2890 from amit-shafnir/worktree-nanoclaw-templates feat(templates): local template loader, ncl --template, and docs	2026-07-02 14:56:45 +03:00
Amit Shafnir	411f5e71df	feat(templates): local template loader, ncl --template, provider-agnostic persona and skills seams Agent templates: folder-only templates under templates/ (context/instructions.md + optional context extras, .mcp.json, skills/). Stamping via ncl groups create --template writes the provider-neutral instructions.prepend.md (inlined at the top of CLAUDE.md/AGENTS.md every spawn), copies context extras preserving their template-relative layout, writes MCP servers to container config, and installs the per-group skills overlay. Includes docs (docs/templates.md, templates/README.md). Setup-wizard wiring ships separately on top of this. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-07-02 12:36:33 +03:00
gavrielc	cb6e3d117c	Merge pull request #2885 from PartridgeNet/fix/slack-setup-socket-mode fix(setup): offer Slack Socket Mode in the guided setup flow	2026-06-30 23:05:45 +03:00
gavrielc	ed7e3f70da	Merge branch 'main' into fix/slack-setup-socket-mode	2026-06-30 23:04:38 +03:00
github-actions[bot]	557e073c2f	chore: bump version to 2.1.23	2026-06-30 19:28:09 +00:00
gavrielc	91ebc9def2	chore(container): bump claude-code, agent SDK to latest - @anthropic-ai/claude-code (cli-tools.json): 2.1.170 → 2.1.197 - @anthropic-ai/claude-agent-sdk: ^0.3.170 → ^0.3.197 - @anthropic-ai/sdk: ^0.100.0 → ^0.108.0 Typecheck and all 112 agent-runner tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-30 22:27:46 +03:00
github-actions[bot]	14c89e9716	docs: update token count to 204k tokens · 102% of context window	2026-06-30 15:49:15 +00:00
github-actions[bot]	549c424a38	chore: bump version to 2.1.22	2026-06-30 15:49:12 +00:00
gavrielc	186b9befcd	Merge pull request #2880 from johnmathews/fix/2828-inbox-symlink-containment-upstream fix(security): contain inbox symlink escapes in attachment writes (#2828)	2026-06-30 18:48:59 +03:00
John Mathews	863d413d32	Merge branch 'main' into fix/2828-inbox-symlink-containment-upstream	2026-06-29 18:27:07 +02:00
Rob Stevenson	cf8478ffbb	fix(setup): offer Slack Socket Mode in the guided setup flow PR #2837 added Slack Socket Mode end-to-end ("adapter + guided setup") but was merged into the `channels` branch, not `main`. As a result the `setup:auto` Slack flow on main is webhook-only: it always collects a signing secret and pushes the user toward a public Request URL, with no Socket Mode option — even though setup/verify.ts already recognizes SLACK_APP_TOKEN and the channels-branch adapter supports it. Forward-port the setup-side of #2837 onto current main, re-authored on top of main's current flow (back-nav, inline agent wiring, operator-role prompt, welcome DM all preserved): - setup/channels/slack.ts: add a mode picker (askSlackMode), mode-specific app-creation steps, collectAppToken() for the xapp- app-level token, conditional credential collection + env, and a mode-aware post-install checklist (Socket Mode skips the public-URL guidance). - setup/add-slack.sh: require/persist either SLACK_APP_TOKEN (Socket Mode) or SLACK_SIGNING_SECRET (webhook) instead of mandating the signing secret. Scope is setup-side only: the adapter's socket support already lives on `channels` (#2837/#2839) and reaches users via /add-slack; the add-slack SKILL.md doc is handled by #2700.	2026-06-29 14:45:31 +01:00
github-actions[bot]	8be5be93ba	docs: update token count to 203k tokens · 101% of context window	2026-06-29 05:58:05 +00:00
omri-maya	add3fc8f70	Merge pull request #2882 from nanocoai/fix/ncl-messaging-group-instance fix(ncl): default messaging-groups create instance to channel_type	2026-06-29 08:57:53 +03:00
Omri Maya	0d841bcd05	fix(ncl): default messaging-groups create instance to channel_type `ncl messaging-groups create` failed with a NOT NULL violation on the `instance` column (migration 016). The generic CRUD insert builds its column list from the resource definition, and `instance` wasn't declared there — so the INSERT omitted the column entirely. The router path never hit this because it goes through `createMessagingGroup`, which has its own `instance ?? channel_type` fallback. There is no operator-facing reason to require `--instance`: the default instance IS the channel type (migration 016, `createMessagingGroup`, and the default-instance resolver all encode this). So rather than force a flag, default it. - crud: add `defaultFrom` to ColumnDef — default a column to another already-resolved column's value on create. Generic, reusable. - messaging-groups: declare `instance` with `defaultFrom: 'channel_type'` (placed after channel_type so it resolves first), still overridable via `--instance` for multi-instance setups. - test: drive the real dispatch('messaging-groups-create') path; asserts omitted -> channel_type and explicit --instance preserved. Goes red if the column/defaultFrom wiring is deleted (insert fails NOT NULL). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-29 08:38:22 +03:00
John Mathews	dd1d0e5677	fix(security): contain channel-inbound attachments via shared inbox guard (#2828 ) extractAttachmentFiles (the channel-inbound attachment path) hardened only the per-message inbox subdir, not the `inbox` root itself. A compromised container can write inside its own session dir, so it could replace `inbox` with a symlink: mkdirSync({recursive}) then followed it, and the realpath containment check passed because it compared against realpathSync(inboxRoot) — which had already followed the symlink. A brand-new attachment file (the `wx` flag only blocks an existing dst) therefore landed outside the session sandbox. This is the same symlink-follow class fixed for the A2A path in #2828 (CWE-59), but reachable from ordinary inbound messages. Extract the guard both inbound paths need into src/inbox-safety.ts (ensureContainedInboxDir + isPathInside): lstat-reject a pre-placed symlink or non-dir at the inbox root AND the per-message subdir before mkdir, then a realpath containment check. forwardAttachedFiles and extractAttachmentFiles now share it, removing the duplicated guard logic. Callers still write with an exclusive flag (COPYFILE_EXCL / wx) and skip-with-warn on failure. Adds src/session-manager.attachments.test.ts (red before this change, green after) covering the symlinked inbox-root vector on the channel path.	2026-06-29 00:27:58 +02:00
John Mathews	36afa40857	fix(agent-to-agent): containment-check target inbox in forwardAttachedFiles (#2828 ) forwardAttachedFiles hardened only the source side of A2A attachment forwarding; the target side called fs.mkdirSync({recursive}) and fs.copyFileSync without any symlink or containment checks. A compromised target agent that can write inside its own session dir could pre-place `inbox` (or `inbox/<future-msgId>`) as a symlink pointing anywhere host-writable — mkdirSync silently follows it and copyFileSync lands attacker-influenced bytes outside the sandbox (CWE-59, GHSA #2828). This mirrors the existing defensive pattern in src/session-manager.ts saveAttachments(): lstat-reject a pre-existing symlink/non-dir at the inbox root and the per-message subdir before mkdir, realpath + isPathInside containment check, and an exclusive (COPYFILE_EXCL) copy that refuses to follow or overwrite a pre-placed symlinked destination. Failures log.warn with structured context and skip rather than throw, so one bad attachment never kills a batch. Tests cover a symlinked inbox dir, a symlinked inbox/<msgId> subdir, a pre-existing symlinked destination file, and a normal end-to-end forward regression.	2026-06-29 00:27:58 +02:00
gavrielc	2afbd18233	Merge pull request #2859 from cben0ist/fix/migrate-v2-is-main fix(migrate-v2): don't SELECT is_main from v1 registered_groups	2026-06-26 13:38:42 +03:00
gavrielc	953496dc37	Merge branch 'main' into fix/migrate-v2-is-main	2026-06-26 13:38:27 +03:00
Christophe Benoist	797491d8b3	fix(migrate-v2): don't SELECT is_main from v1 registered_groups The v2 DB seed queried `is_main` from the v1 `registered_groups` table, but that column was a later v1 addition — older v1 installs (e.g. 1.1.0) don't have it, so the migration's `1b-db` step crashes with `no such column: is_main` and v2.db is never created, cascading into the sessions and tasks steps failing. `is_main` was selected into the V1Group interface but never read anywhere, so this just drops it from the SELECT and the interface. The accompanying comment already states the intent ("Query only the columns we know exist in all v1 installs") — the code now matches it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-25 15:43:19 -04:00
github-actions[bot]	2df754459b	chore: bump version to 2.1.21	2026-06-25 18:59:54 +00:00
gavrielc	0896d4089e	Merge pull request #2832 from nanocoai/feat/reject-with-reason feat(approvals): reject with reason	2026-06-25 21:59:42 +03:00
gavrielc	d153d91307	Merge branch 'main' into feat/reject-with-reason	2026-06-25 21:45:12 +03:00
gavrielc	ce55af12d5	Merge pull request #2843 from robbyczgw-cla/feat/learn-skill feat: add /learn skill — distill or refine a reusable skill from anything	2026-06-25 21:41:23 +03:00
gavrielc	545800a94e	Merge branch 'main' into feat/learn-skill	2026-06-25 21:40:55 +03:00
github-actions[bot]	bfb309bd0c	chore: bump version to 2.1.20	2026-06-25 18:34:34 +00:00
gavrielc	38d9390eea	Merge pull request #2856 from nanocoai/container-limits feat(container): per-container CPU/memory limits (opt-in)	2026-06-25 21:34:16 +03:00
robbyczgw-cla	520ec44aec	feat: add /learn skill — distill or refine a reusable skill from anything Instruction-only skill that distills a reusable skill from any source (directory, URL, pasted notes, or the current conversation) or refines an existing skill in place. Uses existing agent tools (Read/Grep/Glob/WebFetch/ Write) and injects the project's skill-authoring guidelines. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-24 09:41:49 +02:00
gavrielc	8c6a243ffd	Merge branch 'main' into feat/reject-with-reason	2026-06-23 15:46:02 +03:00
Moshe Krupper	2ac7809385	feat(agent-to-agent): clarify the a2a gate approval prompt Replace the terse "Approve delivery?" with a one-line legend that names all three buttons and notes that "Reject with reason…" prompts the approver to type a reason relayed back to the sender. The longer line also widens the card bubble, easing the three-button-row truncation on platforms that size buttons to the message width. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-22 17:16:03 +03:00
Moshe Krupper	e8148bc0a7	feat(approvals): reject-with-reason — relay an optional decline reason to the agent Add a third "Reject with reason…" button to module approval cards. Plain Reject stays the instant fast path; the new option holds the row (status='awaiting_reason'), DM-prompts the approver, and captures their next DM (≤280 chars, truncated) as a one-line reason relayed to the requesting agent as a single combined message. A ghosted hold is finalized as a plain reject by the host sweep after ~5 min — restart-safe via the durable DB row. - Generalize the router message-interceptor to a list (registerMessageInterceptor) so approvals can capture replies alongside the permissions agent-naming flow. - Share reject finalization across the instant, captured, and swept paths via finalizeReject. - Scope: all module approvals (create_agent, install_packages, add_mcp_server); OneCLI credential cards are unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-21 14:55:01 +03:00