Debug
Memory-safe engineering-peer plugin for Claude Code + Codex: 11 skills, 6 bounded agents, 4 hooks. Claims only what it can prove; closes every turn with one next action.
npx -y skills add menot-you/n-soul --skill debugAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 15 days oldThe repository was created 15 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 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
Isolate a failure with a reliable reproducer, differentiate ranked causes, remove the root cause, and prove recovery.
SKILL.md
1.9 KB, 384 tokens by cl100k_base, as published. Nobody here has run it
Debug
Use for failures, regressions, flaky behavior, or any task after one missed fix. After two missed fixes, this method is mandatory before another patch. Keep the engineer focused on the failure mechanism, differentiating experiment, and proof.
Inputs to lock
- User-visible failure and required behavior.
- Exact environment, input, and last known good state.
- Current logs, traces, fixtures, and commands.
- Layers crossed from input to failure.
- Safe boundaries for experiments.
Method
- Reproduce the failure reliably. Record the command, input, output, and frequency.
- Map the failure chain across caller, application, service, host, network, datastore, and external dependency.
- Rank candidate causes by proof. For each one, state what observation would reject it.
- Design the smallest experiment that separates the leading candidates. Change one variable at a time.
- Gather proof before editing. Do not convert an error label into a cause without the mechanism.
- Reduce the reproducer until it still fails and irrelevant layers are removed.
- State the root cause as input, broken mechanism, and user impact.
- Add a regression test that is red on the broken state.
- Apply the smallest clean fix at the owning layer. Avoid retries or fallbacks that conceal the defect.
- Prove the regression test green, run affected broader checks, and retest the original path.
- Verify rollback and watch for adjacent failure modes created by the fix.
Output
Return one focused debug close:
- failure mechanism;
- ranked proof and rejected candidates;
- reproducer;
- fix and why it removes the cause;
- before/after proof;
- remaining uncertainty and one next action if not closed.
If the failure cannot be reproduced, say so and stop at the best proof-gathering action. Do not claim a fix.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.