Claude Code Plugins

Community-maintained marketplace

Feedback

systematic-debugging

@romiluz13/cc10x
8
0

Four-phase debugging discipline that enforces LOG FIRST, hypothesis-driven fixes, and regression tests. Use whenever diagnosing bugs, test failures, or unexpected behaviour.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name systematic-debugging
description Four-phase debugging discipline that enforces LOG FIRST, hypothesis-driven fixes, and regression tests. Use whenever diagnosing bugs, test failures, or unexpected behaviour.
allowed-tools Read, Grep, Glob, Bash

Systematic Debugging

Core Principle

No fixes without root-cause evidence.
If you have not logged or observed the failing state, you are guessing.

Four Phases

  1. Investigate
    • Reproduce the issue reliably.
    • Capture complete error messages, logs, request/response payloads.
    • Inspect recent changes (git log, git diff).
  2. Pattern Analysis
    • Compare failing paths to working examples.
    • Consult the relevant skill guidance (security-patterns, integration-patterns, etc.).
  3. Hypothesis + Minimal Fix
    • State the suspected root cause in one sentence.
    • Write a failing regression test that proves the bug.
    • Apply the smallest change to make the test pass.
  4. Verification
    • Run the regression test and surrounding suite; capture outputs.
    • Summarise the fix, evidence, and follow-up monitoring.

LOG FIRST Playbook

Whenever data shape or external behaviour is unclear:

  • Log the full object/response with JSON.stringify(value, null, 2).
  • Record headers, cookies, environment variables (masking secrets).
  • For HTTP issues, log method, URL, payload, status, and body.
  • For auth, log session claims, roles, and config.

Deliverable Template

## Root Cause
- <what failed and why>

## Evidence
- Logs: <snippet>
- Commands: <command> -> exit <code>

## Fix
- Summary of change.
- Regression test: <file/test name>.

## Verification Summary
- <tests run, exit codes, residual risks>

Anti-Patterns

  • Applying multiple fixes without proving the cause.
  • Trusting documentation over runtime data.
  • Reporting success without logs/tests.

References

  • Official debugging guidance: docs/reference/03-SUBAGENTS.md
  • Verification requirements: plugins/cc10x/skills/verification-before-completion/SKILL.md