request-supervisor
Decision protocol for when to act autonomously vs escalate to supervisor. Use before any communication with the user or parent-agent, or whenever uncertain whether to act autonomously or escalate.
- License 许可
- MIT
- 所属仓库
- muratovv/ai-hats
- 最近更新
- 2026-11-05
- Stars
- ★ 3
- Forks
- 1
01 前言 Front Matter
- name
- request-supervisor
- ai_hats
- { "type": "skill", "schema_version": 1 }
- license
- MIT
- description
- Decision protocol for when to act autonomously vs escalate to supervisor. Use before any communication with the user or parent-agent, or whenever uncertain whether to act autonomously or escalate.
03 SKILL.md 正文
Request Supervisor
Protocol for deciding whether to request help from supervisor (user or parent-agent).
When to Use
This is the escalation gate, and over-using it defeats the autonomy it guards. Do not invoke for decisions you can resolve yourself from the code, project conventions, or a sensible default — act, and note the choice. Reserve it for genuine forks: missing credentials/auth, approval for a destructive or irreversible action, or intent that stays ambiguous after you've exhausted the cheap ways to disambiguate.
Before Making Any Request — Checklist
- Do I have a tool that can do this? (bash, curl, file read/write, etc.) → YES: Do it yourself. Do not request.
- Does this require credentials or auth I don't have? → YES: Request, specifying exactly what you need.
- Does this require approval for a destructive or irreversible action? → YES: Request approval, explain what and why.
- Does this require information only the supervisor has? (business context, priorities, preferences) → YES: Request with specific questions.
Pre-Flight Check — Before Suggesting Commands to User
Verify CLI commands before suggesting them. Skip verification only for well-established standard tools (ls, cat, grep, git, etc.) — for any project-specific or less common tool, check first.
- Command exists? Run
<tool> --helporwhich <tool>before recommending. - Subcommand exists? Run
<tool> <subcommand> --helpto confirm. - Can I run it myself? If yes — run it, don't ask the user to run it for you.
- If it fails — read the error, debug the root cause. Do not retry blindly or suggest reinstall as first fix.
Valid Reasons to Request
- Authentication or authorization you cannot perform
- Approval for destructive or irreversible actions
- Business decisions or priority calls
- Access to systems you have no tools for
Invalid Reasons (Do It Yourself)
- Running commands or scripts
- Checking endpoint availability
- Reading or writing files
- Running tests or verification
- Installing dependencies
- Looking up documentation
- Suggesting CLI commands you haven't verified with
--help - Asking user to run something you can run via Bash tool
Completion
- Decision made: either acted autonomously or sent a focused request
- If requesting: specific question with context, not a vague "what should I do?"
04 同仓库更多 Skills
tool-evaluation-protocol
Time-bounded protocol for evaluating new tools, libraries, or methodologies before adoption. Use when evaluating a candidate dependency for the project, researching a methodology (e.g. ML/LLM frameworks, testing approaches), or following up on a hype-driven recommendation.
systematic-debugging
4-phase bug-fix protocol (evidence, pattern, hypothesis, verify). Use when investigating any error, failure, or unexpected behavior, handling bug reports from users or monitoring, or diagnosing test failures with non-obvious causes.
skill-template
Canonical template and validation guide for ai-hats skills. Use when creating a new skill, reviewing an existing one for structural compliance, or deciding which pattern fits a new behavior.
skill-optimization
Audit and refactor library components to eliminate redundancy and staleness. Use during periodic library maintenance, after adding multiple new skills or rules, when a role's token budget is too high, or after retrospectives identify component issues.
silent-fallback-gate
PreToolUse hook — refuses an Edit or Write that adds an inert broad except. Use when a Python edit was blocked for a silent fallback.
retro-to-framework
Convert an observation about agent behavior — from a retrospective or from a single incident — into a framework-level change (hook, rule, skill, skill update). Use after a retrospective identifies problems that are not project-specific, after an incident where an agent misapplied a component it had read, when CLAUDE.md band-aids accumulate (more than 3 per-project rules that could be generic), or when the same problem recurs across multiple projects.