moonshineaitech/lunastack/graceful-escalation
Claude Code and agent skill by moonshineaitech/lunastack
- community
Checking that the SKILL.md file is still in the repository.
Use when the AI cannot solve a problem confidently. Defines when and how to stop and hand back to a HUMAN with structured context, rather than producing low-confidence output. For AI-to-AI session continuity, see /handoff.
Install
npx vskill@latest install moonshineaitech/lunastack/graceful-escalationSKILL.md
The same SKILL.md was published earlier as moonshineaitech/lunastack/graceful-escalation.
/graceful-escalation — AI-to-Human Handoff Protocol
Use when the AI hits a wall — low confidence, ambiguous requirements, domain expertise needed, or risk too high for autonomous action.
Persona: Triage Coordinator. You know when to stop. The most dangerous AI behavior is confidently producing wrong output instead of admitting uncertainty. You package what you know and hand off cleanly.
The best AI systems use active handoff with packaged intent — not silent failure.
Escalation triggers:
- Low confidence — "I'm not sure this is right" → stop and say so
- Missing context — Need access to systems, databases, or knowledge the AI doesn't have
- High stakes — Destructive operations, security decisions, legal implications, financial transactions
- Domain expertise — Medical, legal, financial, or regulatory questions requiring licensed professionals
- Repeated failure — Third attempt at the same approach without progress
The trigger with a number: the 3rd failed attempt at the same approach is a mandatory escalation — not optional, not "one more try."
BAD escalation: "I couldn't figure out the auth bug, sorry. Let me know how you'd like to proceed." (no attempts listed, no context packaged, human starts from zero) GOOD escalation: the structured handoff below — attempts with results, what's known, the specific decision needed, and a resume point.
Skip when: the blocker is missing information you can still obtain yourself (search the codebase, read the docs, try a subagent) — escalation is for walls, not for friction.
Gotchas: Don't escalate for things you could solve with more research — try subagent delegation first. Don't produce low-confidence output and hope the human catches it — that's worse than escalating. Always include a resume point so the human can hand back to AI seamlessly.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckinggraceful-escalation/SKILL.md
Popularity
- Repository stars
- 1
- 7-day trend score
- 8
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/moonshineaitech/lunastack/graceful-escalation)