メニュー
Inspect, search, and message between conversation sessions. Use when: the user asks for chat history, wants to send a message into another session, or you need to spawn a sub-agent. NOT for: short in-process delegation (use llm_task) or file-backed memory (use memory-tools).
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | sessions |
| description | Inspect, search, and message between conversation sessions. Use when: the user asks for chat history, wants to send a message into another session, or you need to spawn a sub-agent. NOT for: short in-process delegation (use llm_task) or file-backed memory (use memory-tools). |
| metadata | {"all_agents":{"emoji":"💬"}} |
Each chat lives under a session id; transcripts are stored on disk under
~/.all-agents/sessions/. Three primitives expose them:
sessions_list { } — list session ids known to the runtime.sessions_history { session_id, limit } — read the last N messages.sessions_send { session_id, content, role } — append a message into
the named session as if it had been typed (default role: user).Plus two ways to start work in a different agent:
llm_task { prompt, agent } — synchronous in-process delegation
(cheap, no isolation).subagent_spawn { agent, message, timeout } — spawn the target agent
in its own OS subprocess and wait for the reply (isolated, slower).| Need | Use |
|---|---|
| Quick summarisation / classification | llm_task |
| Untrusted input / heavy tool use | subagent_spawn |
| Inject a follow-up into a parallel chat | sessions_send |
| Audit what was said | sessions_history |
subagent_spawn already tears down the child process when the call
returns. You don't need to call processes stop afterwards.
Set up and use 1Password CLI (op). Use when installing the CLI, enabling desktop app integration, signing in (single or multi-account), or reading/injecting/running secrets via op.
Apply a unified diff to the workspace. Use when: the user provides a patch, you generated a diff and want to apply it, or you're integrating an upstream change. NOT for: small surgical edits to a known file (use file-ops write_file).
Monitor blogs and RSS/Atom feeds for updates using the blogwatcher CLI.
Drive a real browser via Playwright for tasks that need JavaScript execution, login forms, or interactive UI. Use when: scraping a SPA, filling out a form, taking a screenshot, downloading from a JS-protected page, or running an end-to-end check. NOT for: static HTML (use web_fetch), public APIs (use curl), or anything where a simple HTTP GET works.
Use the ClawHub CLI to search, install, update, and publish agent skills from clawhub.com. Use when you need to fetch new skills on the fly, sync installed skills to latest or a specific version, or publish new/updated skill folders with the npm-installed clawhub CLI.
Delegate coding tasks (build features, refactor, batch PR reviews, parallel issue fixes via Codex/Claude Code/Pi/OpenCode) by spawning the `coding-agent` sub-agent. Use when: (1) the work needs file exploration + multi-step edits, (2) you want a separate process so its tool noise stays out of your context, (3) you need parallel coding sessions. NOT for: simple one-liner edits (just edit), reading code (use read tools), thread-bound ACP requests (use sessions).