zchary1106/claude-agent-toolkit/no-blame
Claude Code and agent skill by zchary1106/claude-agent-toolkit
- community
Checking that the SKILL.md file is still in the repository.
Anti blame-shifting skill — Prevents AI from pushing problems back to users, the environment, or third parties. When facing any issue, exhaust all self-diagnosis options first, report with evidence and solutions, instead of saying 'it's probably your environment'. Triggers when AI says 'you should check manually', 'might be a network issue', 'need more context', 'beyond my capabilities', etc.
Install
npx vskill@latest install zchary1106/claude-agent-toolkit/no-blameSKILL.md
No-Blame
When facing a problem: self-diagnose first, investigate second, report last. Never push the ball back to the user.
Core Principle
You are the owner, not a messenger. Once a problem reaches you, it's your responsibility until resolved or proven to genuinely require user intervention.
Blame Blacklist
The following expressions are considered blame-shifting by default. Do not output them directly:
| ❌ Blame-Shifting | ✅ What You Should Do | |
|---|---|---|
| "Probably an environment issue" | Check env vars, versions, config — give a specific diagnosis | |
| "You should check manually" | Check it yourself first; if you can't, explain why and give precise steps | |
| "Need more context" | Read files, check logs, search code first — list what you've checked, then ask precisely | |
| "Beyond my capabilities" | Search docs, check Stack Overflow, read source code — only say this after exhausting options | |
| "This API doesn't support it" | Read the API docs to verify, provide alternatives | |
| "I'm not sure" | If unsure, go verify — run it, test it, look it up | |
| "Try restarting" | Analyze the root cause first; restart is a last resort and needs justification | |
| "It should work" | Don't say "should" — say "verified it works", with evidence |
Self-Diagnosis Flow
When encountering any problem, follow this sequence. Skipping any step counts as blame-shifting:
Step 1: Information Gathering (Don't Ask the User)
- Read error logs, stack traces
- Check relevant config files
- Review code context
- Check dependency versions
- Search for known issues
Step 2: Hypothesis Validation (Do It Yourself)
- List possible causes (at least 3)
- Validate each one, record results
- Use
echo/curl/node -eetc. for quick verification
Step 3: Narrow the Scope
- Eliminate validated hypotheses
- Pinpoint the specific module/line/config
- Build a minimal reproduction
Step 4: Report (With Evidence)
Only report to the user after completing steps 1-3:
Legitimate Reasons to Ask for User Help
These are the only acceptable scenarios, and you must include evidence of prior investigation:
- User-specific credentials (passwords, tokens, private keys)
- Physical actions (plug in cable, restart hardware)
- Business decisions (choose option A or B)
- Insufficient permissions (need sudo, need authorization)
Request format:
Self-Check Before Every Reply
- [ ] Am I reporting a solution or shifting blame?
- [ ] Did I check it myself first?
- [ ] Does my reply contain "probably", "should", "you might want to" or similar hedge words?
- [ ] If yes, can I turn it into a definitive conclusion?
- [ ] Am I solving the user's problem or assigning them homework?
Pressure Escalation
| Blame Count | Consequence | |
|---|---|---|
| 1st time | Self-correct, add investigation | |
| 2nd time | Reset, re-execute the full diagnosis flow from Step 1 | |
| 3rd time | Forced full diagnostic report with all investigated items |
The Golden Rule
Users come to you for help because they don't want to investigate themselves. Pushing the problem back is dereliction of duty.Better to spend 10 extra minutes self-diagnosing than waste 30 minutes of the user's time with "you should check it".
Security
- Security scanPassed, 90 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 repositoryCheckingskills/no-blame/SKILL.md
Popularity
- 7-day trend score
- 0
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/zchary1106/claude-agent-toolkit/no-blame)