Merge branch 'main' into cleanup/command-gate-start-and-failopen

feat(approvals): render OneCLI approval requests from the gateway's structured summary

container/agent-runner/src/providers/index.ts

@@ -3,4 +3,3 @@
 // level. Skills add a new provider by appending one import line below.
 
 import './claude.js';
-import './mock.js';

docs/architecture-diagram.md

@@ -31,7 +31,7 @@ flowchart TB
   subgraph Session["Per-Session Container (Docker / Apple Container)"]
     direction TB
     PollLoop["Poll Loop<br/>(container/agent-runner)"]
-    Provider["Agent providers<br/>(claude, opencode, mock; todo: codex)"]
+    Provider["Agent providers<br/>(claude, opencode; todo: codex)"]
     MCP["MCP Tools<br/>send_message, send_file, edit_message,<br/>add_reaction, send_card, ask_user_question,<br/>schedule_task, create_agent,<br/>install_packages, add_mcp_server"]
     Skills["Container Skills<br/>(container/skills/)"]
     InDB[("inbound.db<br/>host writes<br/>even seq<br/>messages_in<br/>destinations<br/>processing_ack")]

groups/global/CLAUDE.md

@@ -1,166 +0,0 @@
-# Main
-
-You are Main, a personal assistant. You help with tasks, answer questions, and can schedule reminders.
-
-## What You Can Do
-
-- Answer questions and have conversations
-- Search the web and fetch content from URLs
-- **Browse the web** with `agent-browser` — open pages, click, fill forms, take screenshots, extract data (run `agent-browser open <url>` to start, then `agent-browser snapshot -i` to see interactive elements)
-- Read and write files in your workspace
-- Run bash commands in your sandbox
-- Schedule tasks to run later or on a recurring basis
-- Send messages back to the chat
-
-## Communication
-
-Be concise — every message costs the reader's attention.
-
-### Destinations
-
-Each turn, your system prompt lists the destinations available to you. If you only have one destination, just write your response directly — it goes there automatically. If you have multiple, wrap each message in a `<message to="name">...</message>` block:
-
-```
-<message to="family">On my way home, 15 minutes</message>
-<message to="worker-1">kick off the pipeline</message>
-```
-
-Inbound messages are labeled with `from="name"` so you can tell which destination they came from and reply using that same name.
-
-### Mid-turn updates
-
-Use the `mcp__nanoclaw__send_message` tool to send a message mid-work (before your final output). If you have one destination, `to` is optional; with multiple, specify it. Pace your updates to the length of the work:
-
-- **Short work (a few seconds, ≤2 quick tool calls):** Don't narrate. Just do it and put the result in your final response.
-- **Longer work (many tool calls, web searches, installs, sub-agents):** Send a short acknowledgment right away ("On it — checking the logs now") so the user knows you got the message.
-- **Long-running work (many minutes, multi-step tasks):** Send periodic updates at natural milestones, and especially **before** slow operations like spinning up an explore sub-agent, downloading large files, or installing packages.
-
-**Never narrate micro-steps.** "I'm going to read the file now… okay, I'm reading it… now I'm parsing it…" is noise. Updates should mark meaningful transitions, not every tool call.
-
-**Outcomes, not play-by-play.** When the work is done, the final message should be about the result, not a transcript of what you did.
-
-### Internal thoughts
-
-Wrap reasoning in `<internal>...</internal>` tags to mark it as scratchpad — logged but not sent. With multiple destinations, any text outside of `<message>` blocks is also treated as scratchpad. With a single destination, only explicit `<internal>` tags are scratchpad; the rest of your response is sent.
-
-```
-<internal>Compiled all three reports, ready to summarize.</internal>
-
-Here are the key findings from the research…
-```
-
-### Sub-agents and teammates
-
-When working as a sub-agent or teammate, only use `send_message` if instructed to by the main agent.
-
-## Your Workspace
-
-Files you create are saved in `/workspace/group/`. Use this for notes, research, or anything that should persist.
-
-## Memory
-
-The `conversations/` folder contains searchable history of past conversations. Use this to recall context from previous sessions.
-
-When you learn something important:
-- Create files for structured data (e.g., `customers.md`, `preferences.md`)
-- Split files larger than 500 lines into folders
-- Keep an index in your memory for the files you create
-
-## Message Formatting
-
-Format messages based on the channel you're responding to. Check your group folder name:
-
-### Slack channels (folder starts with `slack_`)
-
-Use Slack mrkdwn syntax. Run `/slack-formatting` for the full reference. Key rules:
-- `*bold*` (single asterisks)
-- `_italic_` (underscores)
-- `<https://url|link text>` for links (NOT `[text](url)`)
-- `•` bullets (no numbered lists)
-- `:emoji:` shortcodes
-- `>` for block quotes
-- No `##` headings — use `*Bold text*` instead
-
-### WhatsApp/Telegram channels (folder starts with `whatsapp_` or `telegram_`)
-
-- `*bold*` (single asterisks, NEVER **double**)
-- `_italic_` (underscores)
-- `•` bullet points
-- ` ``` ` code blocks
-
-No `##` headings. No `[links](url)`. No `**double stars**`.
-
-### Discord channels (folder starts with `discord_`)
-
-Standard Markdown works: `**bold**`, `*italic*`, `[links](url)`, `# headings`.
-
----
-
-## Installing Packages & Tools
-
-Your container is ephemeral — anything installed via `apt-get` or `pnpm install -g` is lost on restart. To install packages that persist, use the self-modification tools:
-
-1. **`install_packages`** — request system (apt) or global npm packages. Requires admin approval.
-2. **`request_rebuild`** — rebuild your container image so approved packages are baked in. Always call this after `install_packages` to apply the changes.
-
-Example flow:
-```
-install_packages({ apt: ["ffmpeg"], npm: ["@xenova/transformers"], reason: "Audio transcription" })
-# → Admin gets an approval card → approves
-request_rebuild({ reason: "Apply ffmpeg + transformers" })
-# → Admin approves → image rebuilt with the packages
-```
-
-**When to use this vs workspace pnpm install:**
-- `pnpm install` in `/workspace/agent/` persists on disk (it's mounted) but isn't on the global PATH — use it for project-level dependencies
-- `install_packages` is for system tools (ffmpeg, imagemagick) and global npm packages that need to be on PATH
-
-### MCP Servers
-
-Use **`add_mcp_server`** to add an MCP server to your configuration, then **`request_rebuild`** to apply. Browse available servers at https://mcp.so — it's a curated directory of high-quality MCP servers. Most Node.js servers run via `pnpm dlx`, e.g.:
-
-```
-add_mcp_server({ name: "memory", command: "pnpm", args: ["dlx", "@modelcontextprotocol/server-memory"] })
-request_rebuild({ reason: "Add memory MCP server" })
-```
-
-## Task Scripts
-
-For any recurring task, use `schedule_task`. This is the scheduling path — tasks persist across sessions and restarts, and support the pre-task `script` hook described below. Other scheduling tools you might discover (e.g. `CronCreate`, `ScheduleWakeup`) are session-scoped SDK builtins and won't behave the way NanoClaw users expect, so stick with `schedule_task`.
-
-To inspect or change existing tasks, use `list_tasks` (returns one row per series with the stable id) and `update_task` / `cancel_task` / `pause_task` / `resume_task`. Prefer `update_task` over cancel + reschedule — it preserves the series id the user already knows.
-
-Frequent agent invocations — especially multiple times a day — consume API credits and can risk account restrictions. If a simple check can determine whether action is needed, add a `script` — it runs first, and the agent is only called when the check passes. This keeps invocations to a minimum.
-
-### How it works
-
-1. You provide a bash `script` alongside the `prompt` when scheduling
-2. When the task fires, the script runs first (30-second timeout)
-3. Script prints JSON to stdout: `{ "wakeAgent": true/false, "data": {...} }`
-4. If `wakeAgent: false` — nothing happens, task waits for next run
-5. If `wakeAgent: true` — you wake up and receive the script's data + prompt
-
-### Always test your script first
-
-Before scheduling, run the script in your sandbox to verify it works:
-
-```bash
-bash -c 'node --input-type=module -e "
-  const r = await fetch(\"https://api.github.com/repos/owner/repo/pulls?state=open\");
-  const prs = await r.json();
-  console.log(JSON.stringify({ wakeAgent: prs.length > 0, data: prs.slice(0, 5) }));
-"'
-```
-
-### When NOT to use scripts
-
-If a task requires your judgment every time (daily briefings, reminders, reports), skip the script — just use a regular prompt.
-
-### Frequent task guidance
-
-If a user wants tasks running more than ~2x daily and a script can't reduce agent wake-ups:
-
-- Explain that each wake-up uses API credits and risks rate limits
-- Suggest restructuring with a script that checks the condition first
-- If the user needs an LLM to evaluate data, suggest using an API key with direct Anthropic API calls inside the script
-- Help the user find the minimum viable frequency

groups/main/CLAUDE.md

@@ -1,312 +0,0 @@
-@./.claude-global.md
-# Main
-
-You are Main, a personal assistant. You help with tasks, answer questions, and can schedule reminders.
-
-## What You Can Do
-
-- Answer questions and have conversations
-- Search the web and fetch content from URLs
-- **Browse the web** with `agent-browser` — open pages, click, fill forms, take screenshots, extract data (run `agent-browser open <url>` to start, then `agent-browser snapshot -i` to see interactive elements)
-- Read and write files in your workspace
-- Run bash commands in your sandbox
-- Schedule tasks to run later or on a recurring basis
-- Send messages back to the chat
-
-## Communication
-
-Your output is sent to the user or group.
-
-You also have `mcp__nanoclaw__send_message` which sends a message immediately while you're still working. This is useful when you want to acknowledge a request before starting longer work.
-
-### Internal thoughts
-
-If part of your output is internal reasoning rather than something for the user, wrap it in `<internal>` tags:
-
-```
-<internal>Compiled all three reports, ready to summarize.</internal>
-
-Here are the key findings from the research...
-```
-
-Text inside `<internal>` tags is logged but not sent to the user. If you've already sent the key information via `send_message`, you can wrap the recap in `<internal>` to avoid sending it again.
-
-### Sub-agents and teammates
-
-When working as a sub-agent or teammate, only use `send_message` if instructed to by the main agent.
-
-## Memory
-
-The `conversations/` folder contains searchable history of past conversations. Use this to recall context from previous sessions.
-
-When you learn something important:
-- Create files for structured data (e.g., `customers.md`, `preferences.md`)
-- Split files larger than 500 lines into folders
-- Keep an index in your memory for the files you create
-
-## Message Formatting
-
-Format messages based on the channel. Check the group folder name prefix:
-
-### Slack channels (folder starts with `slack_`)
-
-Use Slack mrkdwn syntax. Run `/slack-formatting` for the full reference. Key rules:
-- `*bold*` (single asterisks)
-- `_italic_` (underscores)
-- `<https://url|link text>` for links (NOT `[text](url)`)
-- `•` bullets (no numbered lists)
-- `:emoji:` shortcodes like `:white_check_mark:`, `:rocket:`
-- `>` for block quotes
-- No `##` headings — use `*Bold text*` instead
-
-### WhatsApp/Telegram (folder starts with `whatsapp_` or `telegram_`)
-
-- `*bold*` (single asterisks, NEVER **double**)
-- `_italic_` (underscores)
-- `•` bullet points
-- ` ``` ` code blocks
-
-No `##` headings. No `[links](url)`. No `**double stars**`.
-
-### Discord (folder starts with `discord_`)
-
-Standard Markdown: `**bold**`, `*italic*`, `[links](url)`, `# headings`.
-
----
-
-## Admin Context
-
-This is the **main channel**, which has elevated privileges.
-
-## Authentication
-
-Anthropic credentials must be either an API key from console.anthropic.com (`ANTHROPIC_API_KEY`) or a long-lived OAuth token from `claude setup-token` (`CLAUDE_CODE_OAUTH_TOKEN`). Short-lived tokens from the system keychain or `~/.claude/.credentials.json` expire within hours and can cause recurring container 401s. The `/setup` skill walks through this. OneCLI manages credentials (including Anthropic auth) — run `onecli --help`.
-
-## Container Mounts
-
-Main has read-only access to the project, read-write access to the store (SQLite DB), and read-write access to its group folder:
-
-| Container Path | Host Path | Access |
-|----------------|-----------|--------|
-| `/workspace/project` | Project root | read-only |
-| `/workspace/project/store` | `store/` | read-write |
-| `/workspace/group` | `groups/main/` | read-write |
-
-Key paths inside the container:
-- `/workspace/project/store/messages.db` - SQLite database (read-write)
-- `/workspace/project/store/messages.db` (registered_groups table) - Group config
-- `/workspace/project/groups/` - All group folders
-
----
-
-## Managing Groups
-
-### Finding Available Groups
-
-Available groups are provided in `/workspace/ipc/available_groups.json`:
-
-```json
-{
-  "groups": [
-    {
-      "jid": "120363336345536173@g.us",
-      "name": "Family Chat",
-      "lastActivity": "2026-01-31T12:00:00.000Z",
-      "isRegistered": false
-    }
-  ],
-  "lastSync": "2026-01-31T12:00:00.000Z"
-}
-```
-
-Groups are ordered by most recent activity. The list is synced from WhatsApp daily.
-
-If a group the user mentions isn't in the list, request a fresh sync:
-
-```bash
-echo '{"type": "refresh_groups"}' > /workspace/ipc/tasks/refresh_$(date +%s).json
-```
-
-Then wait a moment and re-read `available_groups.json`.
-
-**Fallback**: Query the SQLite database directly:
-
-```bash
-sqlite3 /workspace/project/store/messages.db "
-  SELECT jid, name, last_message_time
-  FROM chats
-  WHERE jid LIKE '%@g.us' AND jid != '__group_sync__'
-  ORDER BY last_message_time DESC
-  LIMIT 10;
-"
-```
-
-### Registered Groups Config
-
-Groups are registered in the SQLite `registered_groups` table:
-
-```json
-{
-  "1234567890-1234567890@g.us": {
-    "name": "Family Chat",
-    "folder": "whatsapp_family-chat",
-    "trigger": "@Andy",
-    "added_at": "2024-01-31T12:00:00.000Z"
-  }
-}
-```
-
-Fields:
-- **Key**: The chat JID (unique identifier — WhatsApp, Telegram, Slack, Discord, etc.)
-- **name**: Display name for the group
-- **folder**: Channel-prefixed folder name under `groups/` for this group's files and memory
-- **trigger**: The trigger word (usually same as global, but could differ)
-- **requiresTrigger**: Whether `@trigger` prefix is needed (default: `true`). Set to `false` for solo/personal chats where all messages should be processed
-- **isMain**: Whether this is the main control group (elevated privileges, no trigger required)
-- **added_at**: ISO timestamp when registered
-
-### Trigger Behavior
-
-- **Main group** (`isMain: true`): No trigger needed — all messages are processed automatically
-- **Groups with `requiresTrigger: false`**: No trigger needed — all messages processed (use for 1-on-1 or solo chats)
-- **Other groups** (default): Messages must start with `@AssistantName` to be processed
-
-### Adding a Group
-
-1. Query the database to find the group's JID
-2. Ask the user whether the group should require a trigger word before registering
-3. Use the `register_group` MCP tool with the JID, name, folder, trigger, and the chosen `requiresTrigger` setting
-4. Optionally include `containerConfig` for additional mounts
-5. The group folder is created automatically: `/workspace/project/groups/{folder-name}/`
-6. Optionally create an initial `CLAUDE.md` for the group
-
-Folder naming convention — channel prefix with underscore separator:
-- WhatsApp "Family Chat" → `whatsapp_family-chat`
-- Telegram "Dev Team" → `telegram_dev-team`
-- Discord "General" → `discord_general`
-- Slack "Engineering" → `slack_engineering`
-- Use lowercase, hyphens for the group name part
-
-#### Adding Additional Directories for a Group
-
-Groups can have extra directories mounted. Add `containerConfig` to their entry:
-
-```json
-{
-  "1234567890@g.us": {
-    "name": "Dev Team",
-    "folder": "dev-team",
-    "trigger": "@Andy",
-    "added_at": "2026-01-31T12:00:00Z",
-    "containerConfig": {
-      "additionalMounts": [
-        {
-          "hostPath": "~/projects/webapp",
-          "containerPath": "webapp",
-          "readonly": false
-        }
-      ]
-    }
-  }
-}
-```
-
-The directory will appear at `/workspace/extra/webapp` in that group's container.
-
-#### Sender Allowlist
-
-After registering a group, explain the sender allowlist feature to the user:
-
-> This group can be configured with a sender allowlist to control who can interact with me. There are two modes:
->
-> - **Trigger mode** (default): Everyone's messages are stored for context, but only allowed senders can trigger me with @{AssistantName}.
-> - **Drop mode**: Messages from non-allowed senders are not stored at all.
->
-> For closed groups with trusted members, I recommend setting up an allow-only list so only specific people can trigger me. Want me to configure that?
-
-If the user wants to set up an allowlist, edit `~/.config/nanoclaw/sender-allowlist.json` on the host:
-
-```json
-{
-  "default": { "allow": "*", "mode": "trigger" },
-  "chats": {
-    "<chat-jid>": {
-      "allow": ["sender-id-1", "sender-id-2"],
-      "mode": "trigger"
-    }
-  },
-  "logDenied": true
-}
-```
-
-Notes:
-- Your own messages (`is_from_me`) explicitly bypass the allowlist in trigger checks. Bot messages are filtered out by the database query before trigger evaluation, so they never reach the allowlist.
-- If the config file doesn't exist or is invalid, all senders are allowed (fail-open)
-- The config file is on the host at `~/.config/nanoclaw/sender-allowlist.json`, not inside the container
-
-### Removing a Group
-
-1. Read `/workspace/project/data/registered_groups.json`
-2. Remove the entry for that group
-3. Write the updated JSON back
-4. The group folder and its files remain (don't delete them)
-
-### Listing Groups
-
-Read `/workspace/project/data/registered_groups.json` and format it nicely.
-
----
-
-## Global Memory
-
-You can read and write to `/workspace/global/CLAUDE.md` for facts that should apply to all groups. Only update global memory when explicitly asked to "remember this globally" or similar.
-
----
-
-## Scheduling for Other Groups
-
-When scheduling tasks for other groups, use the `target_group_jid` parameter with the group's JID from `registered_groups.json`:
-- `schedule_task(prompt: "...", schedule_type: "cron", schedule_value: "0 9 * * 1", target_group_jid: "120363336345536173@g.us")`
-
-The task will run in that group's context with access to their files and memory.
-
----
-
-## Task Scripts
-
-For any recurring task, use `schedule_task`. Frequent agent invocations — especially multiple times a day — consume API credits and can risk account restrictions. If a simple check can determine whether action is needed, add a `script` — it runs first, and the agent is only called when the check passes. This keeps invocations to a minimum.
-
-Use `list_tasks` to see existing tasks (one row per series with the stable id), and `update_task` / `cancel_task` / `pause_task` / `resume_task` to modify them. Prefer `update_task` over cancel + reschedule when adjusting an existing task.
-
-### How it works
-
-1. You provide a bash `script` alongside the `prompt` when scheduling
-2. When the task fires, the script runs first (30-second timeout)
-3. Script prints JSON to stdout: `{ "wakeAgent": true/false, "data": {...} }`
-4. If `wakeAgent: false` — nothing happens, task waits for next run
-5. If `wakeAgent: true` — you wake up and receive the script's data + prompt
-
-### Always test your script first
-
-Before scheduling, run the script in your sandbox to verify it works:
-
-```bash
-bash -c 'node --input-type=module -e "
-  const r = await fetch(\"https://api.github.com/repos/owner/repo/pulls?state=open\");
-  const prs = await r.json();
-  console.log(JSON.stringify({ wakeAgent: prs.length > 0, data: prs.slice(0, 5) }));
-"'
-```
-
-### When NOT to use scripts
-
-If a task requires your judgment every time (daily briefings, reminders, reports), skip the script — just use a regular prompt.
-
-### Frequent task guidance
-
-If a user wants tasks running more than ~2x daily and a script can't reduce agent wake-ups:
-
-- Explain that each wake-up uses API credits and risks rate limits
-- Suggest restructuring with a script that checks the condition first
-- If the user needs an LLM to evaluate data, suggest using an API key with direct Anthropic API calls inside the script
-- Help the user find the minimum viable frequency

package.json

@@ -1,6 +1,6 @@
 {
   "name": "nanoclaw",
-  "version": "2.1.24",
+  "version": "2.1.28",
   "description": "Personal Claude assistant. Lightweight, secure, customizable.",
   "type": "module",
   "packageManager": "pnpm@10.33.0",

repo-tokens/badge.svg

@@ -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="204k tokens, 102% of context window">
-  <title>204k tokens, 102% 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="207k tokens, 104% of context window">
+  <title>207k tokens, 104% 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">204k</text>
-        <text x="71" y="14">204k</text>
+        <text aria-hidden="true" x="71" y="15" fill="#010101" fill-opacity=".3">207k</text>
+        <text x="71" y="14">207k</text>
       </g>
     </g>
   </a>

src/cli/dispatch.test.ts

@@ -2,6 +2,32 @@ import { describe, it, expect, vi, beforeEach } from 'vitest';
 
 // --- Mocks ---
 
+const approvalState = vi.hoisted(() => ({
+  requestApproval: vi.fn(),
+  approvalHandler: null as
+    | null
+    | ((args: {
+        session: unknown;
+        payload: Record<string, unknown>;
+        userId: string;
+        notify: (text: string) => void;
+      }) => Promise<void>),
+  registerApprovalHandler: vi.fn(
+    (
+      action: string,
+      handler: (args: {
+        session: unknown;
+        payload: Record<string, unknown>;
+        userId: string;
+        notify: (text: string) => void;
+      }) => Promise<void>,
+    ) => {
+      if (action === 'cli_command') approvalState.approvalHandler = handler;
+    },
+  ),
+  observedContexts: [] as CallerContext[],
+}));
+
 vi.mock('../log.js', () => ({
   log: { info: vi.fn(), warn: vi.fn(), error: vi.fn(), debug: vi.fn() },
 }));
@@ -29,8 +55,8 @@ vi.mock('./crud.js', () => ({
 }));
 
 vi.mock('../modules/approvals/index.js', () => ({
-  registerApprovalHandler: vi.fn(),
-  requestApproval: vi.fn(),
+  registerApprovalHandler: approvalState.registerApprovalHandler,
+  requestApproval: approvalState.requestApproval,
 }));
 
 // Register a test command so dispatch has something to find
@@ -98,6 +124,18 @@ register({
   handler: async (args) => ({ echo: args }),
 });
 
+register({
+  name: 'approval-context-command',
+  description: 'approval command that records caller context',
+  resource: 'groups',
+  access: 'approval',
+  parseArgs: (raw) => raw,
+  handler: async (_args, ctx) => {
+    approvalState.observedContexts.push(ctx);
+    return { caller: ctx.caller };
+  },
+});
+
 // Commands that return data shaped like real resources (for post-handler filtering tests)
 register({
   name: 'groups-list-data',
@@ -152,6 +190,7 @@ import type { CallerContext } from './frame.js';
 
 beforeEach(() => {
   vi.clearAllMocks();
+  approvalState.observedContexts.length = 0;
   // Default: the four CLI-whitelisted resources with their real scopeFields.
   const scopeFields: Record<string, string> = {
     groups: 'id',
@@ -391,6 +430,39 @@ describe('CLI scope enforcement', () => {
     expect(mockGetContainerConfig).not.toHaveBeenCalled();
   });
 
+  it('approval replay preserves the original agent caller context', async () => {
+    mockGetContainerConfig.mockReturnValue({ cli_scope: 'group' });
+    mockGetSession.mockReturnValue({ id: 's1', agent_group_id: 'g1', messaging_group_id: 'mg1' });
+    mockGetAgentGroup.mockReturnValue({ id: 'g1', name: 'Group One' });
+
+    const ctx = agentCtx();
+    const resp = await dispatch({ id: '1', command: 'approval-context-command', args: {} }, ctx);
+
+    expect(resp.ok).toBe(false);
+    expect(approvalState.requestApproval).toHaveBeenCalledTimes(1);
+
+    const approval = approvalState.requestApproval.mock.calls[0][0] as { payload: Record<string, unknown> };
+    expect(approval.payload).toEqual({
+      frame: {
+        id: '1',
+        command: 'approval-context-command',
+        args: { agent_group_id: 'g1', group: 'g1', id: 'g1' },
+      },
+      callerContext: ctx,
+    });
+
+    expect(approvalState.approvalHandler).toBeTypeOf('function');
+    await approvalState.approvalHandler!({
+      session: { id: 's1', agent_group_id: 'g1', messaging_group_id: 'mg1' },
+      payload: approval.payload,
+      userId: 'telegram:admin',
+      notify: vi.fn(),
+    });
+
+    expect(approvalState.observedContexts).toEqual([ctx]);
+    expect(approvalState.requestApproval).toHaveBeenCalledTimes(1);
+  });
+
   // --- Post-handler filtering ---
 
   it('group: groups list filters out other groups', async () => {

src/cli/dispatch.ts

@@ -14,7 +14,16 @@ import type { CallerContext, ErrorCode, RequestFrame, ResponseFrame } from './fr
 import { getResource } from './crud.js';
 import { lookup } from './registry.js';
 
-export async function dispatch(req: RequestFrame, ctx: CallerContext): Promise<ResponseFrame> {
+type DispatchOptions = {
+  /** True when a command is being replayed after approval. */
+  approved?: boolean;
+};
+
+export async function dispatch(
+  req: RequestFrame,
+  ctx: CallerContext,
+  opts: DispatchOptions = {},
+): Promise<ResponseFrame> {
   let cmd = lookup(req.command);
 
   // Fallback: if the full command isn't registered, trim the last
@@ -101,7 +110,7 @@ export async function dispatch(req: RequestFrame, ctx: CallerContext): Promise<R
     }
   }
 
-  if (ctx.caller !== 'host' && cmd.access === 'approval') {
+  if (ctx.caller !== 'host' && cmd.access === 'approval' && !opts.approved) {
     const session = getSession(ctx.sessionId);
     if (!session) {
       return err(req.id, 'handler-error', 'Session not found.');
@@ -117,7 +126,7 @@ export async function dispatch(req: RequestFrame, ctx: CallerContext): Promise<R
       session,
       agentName,
       action: 'cli_command',
-      payload: { frame: { id: req.id, command: req.command, args: req.args } },
+      payload: { frame: { id: req.id, command: req.command, args: req.args }, callerContext: ctx },
       title: `CLI: ${req.command}`,
       question: `Agent "${agentName}" wants to run:\n\`ncl ${req.command}${argSummary ? ' ' + argSummary : ''}\``,
     });
@@ -178,9 +187,10 @@ export async function dispatch(req: RequestFrame, ctx: CallerContext): Promise<R
   }
 }
 
-registerApprovalHandler('cli_command', async ({ session, payload, userId, notify }) => {
+registerApprovalHandler('cli_command', async ({ payload, notify }) => {
   const frame = payload.frame as RequestFrame;
-  const response = await dispatch(frame, { caller: 'host' });
+  const callerContext = parseCallerContext(payload.callerContext) ?? { caller: 'host' };
+  const response = await dispatch(frame, callerContext, { approved: true });
 
   if (response.ok) {
     const data = typeof response.data === 'string' ? response.data : JSON.stringify(response.data, null, 2);
@@ -190,6 +200,26 @@ registerApprovalHandler('cli_command', async ({ session, payload, userId, notify
   }
 });
 
+function parseCallerContext(value: unknown): CallerContext | undefined {
+  if (!value || typeof value !== 'object') return undefined;
+  const record = value as Record<string, unknown>;
+  if (record.caller === 'host') return { caller: 'host' };
+  if (
+    record.caller === 'agent' &&
+    typeof record.sessionId === 'string' &&
+    typeof record.agentGroupId === 'string' &&
+    typeof record.messagingGroupId === 'string'
+  ) {
+    return {
+      caller: 'agent',
+      sessionId: record.sessionId,
+      agentGroupId: record.agentGroupId,
+      messagingGroupId: record.messagingGroupId,
+    };
+  }
+  return undefined;
+}
+
 function err(id: string, code: ErrorCode, message: string): ResponseFrame {
   return { id, ok: false, error: { code, message } };
 }

src/command-gate.test.ts

@@ -0,0 +1,83 @@
+/**
+ * Tests for the host-side command gate — filtered commands are dropped
+ * before reaching the container, and admin commands are gated against
+ * the user_roles table.
+ */
+import { afterEach, beforeEach, describe, expect, it } from 'vitest';
+
+import { gateCommand } from './command-gate.js';
+import { closeDb, createAgentGroup, initTestDb, runMigrations } from './db/index.js';
+import { createUser } from './modules/permissions/db/users.js';
+import { grantRole } from './modules/permissions/db/user-roles.js';
+
+function now(): string {
+  return new Date().toISOString();
+}
+
+function seedAgentGroup(id: string): void {
+  createAgentGroup({ id, name: id.toUpperCase(), folder: id, agent_provider: null, created_at: now() });
+}
+
+function seedUser(id: string): void {
+  createUser({ id, kind: 'telegram', display_name: null, created_at: now() });
+}
+
+beforeEach(() => {
+  const db = initTestDb();
+  runMigrations(db);
+  seedAgentGroup('ag-1');
+  seedAgentGroup('ag-2');
+});
+
+afterEach(() => {
+  closeDb();
+});
+
+describe('filtered commands', () => {
+  it('drops /start before it reaches the container', () => {
+    expect(gateCommand('/start', 'telegram:1', 'ag-1')).toEqual({ action: 'filter' });
+  });
+
+  it('drops /start regardless of sender', () => {
+    expect(gateCommand('/start', null, 'ag-1')).toEqual({ action: 'filter' });
+  });
+});
+
+describe('admin gating goes through roles', () => {
+  it('denies an admin command from a non-admin user', () => {
+    expect(gateCommand('/clear', 'telegram:nobody', 'ag-1')).toEqual({ action: 'deny', command: '/clear' });
+  });
+
+  it('denies an admin command with no sender', () => {
+    expect(gateCommand('/clear', null, 'ag-1')).toEqual({ action: 'deny', command: '/clear' });
+  });
+
+  it('allows an admin command from an owner', () => {
+    seedUser('telegram:owner');
+    grantRole({ user_id: 'telegram:owner', role: 'owner', agent_group_id: null, granted_by: null, granted_at: now() });
+    expect(gateCommand('/clear', 'telegram:owner', 'ag-1')).toEqual({ action: 'pass' });
+  });
+
+  it('allows an admin command from a scoped admin of the group', () => {
+    seedUser('telegram:admin');
+    grantRole({
+      user_id: 'telegram:admin',
+      role: 'admin',
+      agent_group_id: 'ag-1',
+      granted_by: null,
+      granted_at: now(),
+    });
+    expect(gateCommand('/clear', 'telegram:admin', 'ag-1')).toEqual({ action: 'pass' });
+    expect(gateCommand('/clear', 'telegram:admin', 'ag-2')).toEqual({ action: 'deny', command: '/clear' });
+  });
+});
+
+describe('normal messages pass through', () => {
+  it('passes a plain message', () => {
+    expect(gateCommand('hello there', 'telegram:1', 'ag-1')).toEqual({ action: 'pass' });
+  });
+
+  it('passes an unknown slash command', () => {
+    expect(gateCommand('/whatever', 'telegram:1', 'ag-1')).toEqual({ action: 'pass' });
+  });
+});

src/command-gate.ts

@@ -7,11 +7,11 @@
  *   "Permission denied" response written directly to messages_out
  * - Normal messages: pass through unchanged
  */
-import { getDb, hasTable } from './db/connection.js';
+import { hasAdminPrivilege } from './modules/permissions/db/user-roles.js';
 
 export type GateResult = { action: 'pass' } | { action: 'filter' } | { action: 'deny'; command: string };
 
-const FILTERED_COMMANDS = new Set(['/help', '/login', '/logout', '/doctor', '/config', '/remote-control']);
+const FILTERED_COMMANDS = new Set(['/start', '/help', '/login', '/logout', '/doctor', '/config', '/remote-control']);
 const ADMIN_COMMANDS = new Set(['/clear', '/compact', '/context', '/cost', '/files', '/upload-trace']);
 
 /**
@@ -48,16 +48,5 @@ export function gateCommand(content: string, userId: string | null, agentGroupId
 
 function isAdmin(userId: string | null, agentGroupId: string): boolean {
   if (!userId) return false;
-  if (!hasTable(getDb(), 'user_roles')) return true; // no permissions module = allow all
-  const db = getDb();
-  const row = db
-    .prepare(
-      `SELECT 1 FROM user_roles
-       WHERE user_id = ?
-         AND (role = 'owner' OR role = 'admin')
-         AND (agent_group_id IS NULL OR agent_group_id = ?)
-       LIMIT 1`,
-    )
-    .get(userId, agentGroupId);
-  return row != null;
+  return hasAdminPrivilege(userId, agentGroupId);
 }

src/container-runner.ts

@@ -320,12 +320,6 @@ export function buildMounts(
     mounts.push({ hostPath: fragmentsDir, containerPath: '/workspace/agent/.claude-fragments', readonly: true });
   }
 
-  // Global memory directory — always read-only.
-  const globalDir = path.join(GROUPS_DIR, 'global');
-  if (fs.existsSync(globalDir)) {
-    mounts.push({ hostPath: globalDir, containerPath: '/workspace/global', readonly: true });
-  }
-
   // Shared CLAUDE.md — read-only, imported by the composed entry point via
   // the `.claude-shared.md` symlink inside the group dir.
   const sharedClaudeMd = path.join(process.cwd(), 'container', 'CLAUDE.md');

src/modules/approvals/onecli-approvals.ts

@@ -254,16 +254,46 @@ async function sweepStaleApprovals(): Promise<void> {
   }
 }
 
+/** The hosted gateway's structured request summary — not yet in the SDK's
+ *  ApprovalRequest type (observed on api.onecli.sh, 2026-07): the action being
+ *  performed plus labeled fields (To / Subject / Body for email sends). */
+interface ApprovalSummary {
+  action?: string;
+  details?: { label: string; value: string }[];
+}
+
+const SUMMARY_VALUE_EXCERPT_CHARS = 900;
+
 function buildQuestion(request: ApprovalRequest, agentName: string): string {
-  const lines = [
-    'Credential access request',
-    `Agent: ${agentName}`,
-    '```',
-    `${request.method} ${request.host}${request.path}`,
-    '```',
-  ];
-  if (request.bodyPreview) {
-    lines.push('Body:', '```', request.bodyPreview, '```');
+  const lines = [`*Agent:* ${agentName}`];
+
+  const summary = (request as ApprovalRequest & { summary?: ApprovalSummary }).summary;
+  if (summary?.details?.length) {
+    if (summary.action) lines.push(`*Action:* ${summary.action}`);
+    // A render bug here must never decide the request: handleRequest's catch
+    // returns 'deny', so stay defensive — coerce non-string values instead of
+    // assuming the gateway's shape, and keep the card under Slack's 3000-char
+    // section limit or delivery itself fails.
+    let budget = 2600;
+    for (const { label, value } of summary.details) {
+      const raw = typeof value === 'string' ? value : (JSON.stringify(value) ?? String(value));
+      const cap = Math.min(SUMMARY_VALUE_EXCERPT_CHARS, Math.max(0, budget));
+      if (cap === 0) {
+        lines.push(`_…${summary.details.length} field(s) omitted for length — see the audit payload._`);
+        break;
+      }
+      const v = raw.length > cap ? `${raw.slice(0, cap)}…` : raw;
+      budget -= v.length + String(label).length + 8;
+      // Multi-line values (message bodies) read better fenced; short labeled
+      // fields (To, Subject) inline.
+      if (v.includes('\n')) lines.push(`*${label}:*`, '```', v, '```');
+      else lines.push(`*${label}:* ${v}`);
+    }
+  } else if (request.bodyPreview) {
+    lines.push('```', request.bodyPreview.slice(0, SUMMARY_VALUE_EXCERPT_CHARS * 2), '```');
+    lines.push(`_${request.method} ${request.host}${request.path}_`);
+  } else {
+    lines.push(`_${request.method} ${request.host}${request.path}_`);
   }
   return lines.join('\n');
 }

src/providers/index.ts

@@ -1,6 +1,6 @@
 // Host-side provider container-config barrel.
 // Providers that need host-side container setup (extra mounts, env passthrough,
 // per-session directories) self-register on import. Providers with no host
-// needs (claude, mock) don't appear here.
+// needs (claude) don't appear here.
 //
 // Skills add a new provider by appending one import line below.

src/providers/provider-container-registry.ts

@@ -7,7 +7,7 @@
  * the registered config fn, and merges the returned mounts/env into the spawn
  * args.
  *
- * Providers without host-side needs (e.g. `claude`, `mock`) don't appear in
+ * Providers without host-side needs (e.g. `claude`) don't appear in
  * this registry at all — the lookup returns `undefined` and the spawn path
  * proceeds with only the default mounts and env.
  *