Debug investigator
Custom skills, instructions, and prompt templates for AI coding agents (Claude Code, Kimi, etc.).
npx -y skills add Flowerf19/agents-skills --skill debug-investigatorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Systematically investigate the root cause of a bug, test failure, performance issue, or unexpected behavior before any fix.
SKILL.md
1.9 KB, as published. Nobody here has run it
Investigation only; do not change production code. Hand off to thoughtful-coder after root cause is confirmed.
Iron law
No fix before root cause. This holds even under time pressure, even when a fix looks "obvious", and especially when previous fixes didn't stick. Investigation is complete when one cause-and-effect statement explains the failure and is backed by evidence — not before.
Workflow
- Reproduce reliably and capture the exact error, inputs, and environment.
- Trace data/control flow from the failure back to its source; check recent changes and affected callers.
- Compare with a nearby working path to isolate the broken assumption.
- Test one explicit hypothesis at a time with the smallest probe. Record what each probe confirms or rules out.
- State the root cause, evidence, blast radius, and smallest correction point.
Provide a failing test or minimal reproduction when feasible, then pass it with the root-cause evidence to thoughtful-coder. After the fix, verify both the reproduction and the relevant full test suite.
Red flags — stop and restart the investigation
- "Quick fix for now, refactor later."
- Trying random changes hoping one sticks.
- Bundling multiple edits in one attempt.
- Proposing a fix before tracing data flow.
Circuit breaker
After three failed hypotheses or fix attempts, stop — the architecture may be wrong, not the line of code. Surface this to the user and re-scope before trying another patch.
Environment or timing failures
If evidence points to environment or timing, document the boundary failure and recommend targeted retries, timeouts, or monitoring without claiming certainty beyond the evidence.