wolfego
wolfego/agent-messenger
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/wolfego/agent-messenger --skill debug命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Show available agent-messenger commands. Use when the user says "/am", asks about messaging commands, or wants to know how to communicate with other agents.
Send a message to another agent via agent-messenger. Use when the user says "send message", "#sm", or wants to communicate with Cursor or another agent.
Show the orchestrate workflow overview. Use when the user says "/orchestrate" or asks about the orchestrator/implementer workflow.
基于 SOC 职业分类
正在显示 SKILL.md
| name | debug |
| description | Show the debug workflow overview. Use when the user says "/debug" or asks about the two-agent debug workflow. |
| disable-model-invocation | true |
Read docs/guidance/workflows/debug.md for the full process. If it doesn't exist, this project hasn't started the debug workflow yet — ask the Cursor orchestrator to run #debug first.
Your role (investigator/fixer): Respond to action fields in messages from the orchestrator:
| Action | Your response |
|---|---|
investigate | Run diagnostic tools, check logs, trace code paths per the hypothesis provided. Report raw findings (observed behavior, not conclusions). |
reproduce | Create a minimal reproduction — ideally a failing test. Report repro steps. |
fix | Implement the fix using TDD: write a failing test that reproduces the bug, then fix it, then verify. |
verify-fix | Run full quality gates (lint, typecheck, test, build). Confirm no regressions. |
Report raw findings, not interpretations. If blocked (need access, can't reproduce), report immediately rather than guessing.