菜单
Drives Gantry agent-chat tests through the local Control API session path and captures assistant responses. Use when testing an agent response, permission prompt, skill behavior, or runtime chat flow without needing Telegram UI rendering.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Orchestrate a Gantry plan to a finished PR — Claude writes a goal-prompt doc, delegates implementation stages to the Codex plugin (xhigh + ponytail), verifies and commits between stages, and closes with the autoreview loop. Use when the user asks to implement a plan/goal prompt, "run the goal pipeline", "same as the goal", or hands over a docs/architecture/*-goal-prompt.md to execute.
End-to-end Gantry plan implementation pipeline. Use when the user asks to implement or pursue a plan, goal prompt, feature, fix, refactor, PR update, or "same as the goal" workflow with subagent implementation, ponytail simplicity, no-commentary implementation, autoreview, build, launchctl restart, Knacklabs lead gen smoke, and PR/pipeline closeout.
Gantry self-administration reference for current host runtime operations: CLI commands, settings.yaml runtime settings, .env credential keys, agent management, scheduler tools, memory tools, browser tools, diagnostics, and service control. Use when asked to manage Gantry itself.
Guides Gantry tool, skill, MCP server, browser, local CLI, semantic capability, inventory, and permission lifecycle work. Use when changing durable capability selection, visible attached sources, transient request_permission flows, or agent access review paths.
Use when creating or revising a Gantry product, UX, permission, job, ingress, provider, settings, or agent-flow plan that must be decision-complete with no room for assumptions. Triggers for PM plan, product manager plan, no assumptions, UX flow, user flow, agent flow, acceptance criteria, validation plan, provider-neutral plan, or Surface Impact Matrix.
Drives Gantry tests through the visible Telegram UI with Chrome. Use only when testing Telegram rendering, Telegram buttons, topic routing, or visible Telegram conversation behavior; use agent-chat-test for direct runtime agent-chat tests.
| name | agent-chat-test |
| description | Drives Gantry agent-chat tests through the local Control API session path and captures assistant responses. Use when testing an agent response, permission prompt, skill behavior, or runtime chat flow without needing Telegram UI rendering. |
Use this skill when the goal is to send a test message to a Gantry agent and
inspect the runtime response. Prefer this over telegram-agent-loop unless the
thing being tested is specifically Telegram UI rendering, Telegram buttons, or
visible Telegram topic behavior.
node .codex/scripts/agent_chat_test.mjs --fresh "Ask a short status question and reply in one sentence."
The script sends through Gantry's app/session Control API. This exercises the real session enqueue path, runtime queue, permission handling, and outbound session events without direct Postgres writes.
npm run build
node dist/cli/index.js service restart
node dist/cli/index.js status
node .codex/scripts/agent_chat_test.mjs --fresh "Use the LinkedIn posting skill to draft a harmless test post, then stop before publishing."
Session identifies the app session used for the test.Conversation identifies the app conversation.Accepted message proves the message was durably accepted.node dist/cli/index.js status
tail -n 200 ~/gantry/logs/gantry.log
--conversation-id codex-test-name to reuse a stable test conversation.--fresh to avoid stale session context by appending a timestamp.--thread-id <id> only when testing thread-scoped behavior.--json when another script needs structured output.--timeout-ms <ms> to shorten or lengthen the wait, up to 300000 ms.telegram-agent-loop for those cases.