Common DebuggingSAFE

A collection of Agent Skills Standard and Best Practice for Programming Languages, Frameworks that help our AI Agent follow best practies on frameworks and programming laguages

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: common-debugging
description: Troubleshoot systematically using the Scientific Method. Use when debugging crashes, tracing errors, diagnosing unexpected behavior, or investigating exceptions.
metadata:
  triggers:
    keywords:
    - debug
    - fix bug
    - crash
    - error
    - exception
    - troubleshooting
---
# Debugging Expert

## **Priority: P1 (HIGH)**

## Root-Cause Protocol

1. **OBSERVE**: Gather error, logs, repro steps, recent diffs.
2. **REPRODUCE**: Make the failure happen on demand or gather more evidence.
3. **HYPOTHESIZE**: State one hypothesis in plain language.
4. **EXPERIMENT**: Change one variable to prove or kill the theory.
5. **FIX**: Touch code only after root cause proven.
6. **VERIFY**: Re-run the failing case and regression checks.

## Red Flags

- **Stop if you are changing code before repro**: You are guessing.
- **Stop if fix #2 starts before understanding fix #1**: Re-open root cause.
- **Stop if "quick patch for now" appears**: Symptom masking starts there.

## Rationalization Prevention

- **"The bug is obvious"**: Obvious bugs still need evidence.
- **"I already tried a few things"**: That means you need structure.
- **"It only fails in prod"**: Gather prod evidence, do not invent local myths.
- **Minimal repro first**: A minimal reproduction beats more random fixes.

## Anti-Patterns

- **No shotgun debugging**: Prove root cause before changing code.
- **No debug prints in production**: Remove all print/console.log before commit.
- **No symptom masking**: Fix root cause; never swallow errors without handling.

## References

- [Bug Report Template](references/bug-report-template.md)

## Incident pressure checklist

- For production incidents, reproduce with a production-like replay and state one hypothesis before changing code; stop symptom patches until the root cause is proven.

## Canonical response anchors

When this skill applies, preserve the following domain terminology or equivalent concrete examples in the answer when relevant:
- hypothesis
- one hypothesis
- stop

- Additional task-grounded exact anchors: minimal

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.