anti-panic-protocol
由 noelistired 提供
Web & Frontend Development简介
Handle tool failures under pressure with bounded retries, clean user communication, and safe escalation.
描述
该技能的功能
Defines a strict error-handling workflow for agents: identify the failure, attempt two bounded fixes, then escalate with plain-language communication. Prevents retry loops, duplicate outbound messages, and raw error dumps reaching the user. Includes hard stops for destructive or security-boundary situations.
Enforces a two-attempt ceiling on retries and requires verified outcomes before reporting success, preventing runaway loops and false "done" confirmations that other ad-hoc error handling skips.
何时使用
- Recovering from a rate-limited publish without spamming retries
- Handling a calendar event creation error with date/timezone fixes
- Escalating a timed-out login flow to the user for OAuth confirmation
- Preventing duplicate posts when an external publish action fails
- Reporting a failed file write as pending rather than falsely done
示例工作流
User asks: Post my announcement to the blog
- 1
Attempt to publish the post via the configured tool
- 2
Receive rate-limit rejection, identify failure in one sentence
- 3
Retry immediately as fix #1
- 4
Retry with a delayed slot as fix #2
- 5
Stop retrying, send escalation message: what failed, what was tried, what happens next
Dit faalde: publish werd afgewezen door rate-limit. Ik heb 1) direct retry en 2) delayed retry geprobeerd. Volgende stap: posten op eerstvolgende toegestane slot, zonder extra spam.
安装
clawhub install noelistired/anti-panic-protocol运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。