gjalla-debug — Claude Code skill: what it does, how to install
gjalla-debug — a skill that guides agents to find the root cause of a bug with evidence before proposing a fix. Works with Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI and OpenCode.
no license 357 installs ★ 3 updated 2026-09-20
npx skills add gjalla/engineering --skill gjalla-debug -a claude-codeWhat gjalla-debug does
gjalla-debug trains the agent to run an investigation: record the exact symptom, reproduce the problem or state why it could not be reproduced, confirm the code is current, pull context from repo rules and change history, locate the root instead of the first observable consequence. After diagnosis, the skill requires counting how many places in the codebase contain the same defect and writing a test that currently fails and will pass after the fix.
The output is a structured report: symptom, mechanism with file and line, why the broken state was reachable, blast radius (list of locations and count), confidence level (what was verified directly, what was inferred), proposed fix with test. If the fix touches system primitives, the skill stops and hands off to gjalla-spec.
The skill forbids forming a theory before gathering data. Without reproduction, it switches to logs, telemetry and database state but marks the diagnosis as less certain.
What it helps with
- locate the root cause instead of the first visible consequence
- reproduce the issue or state why reproduction failed
- count how many locations in the repo contain the same defect
- write a test that currently fails and passes after the fix
- pull context from repo rules and change history
Example prompts for the agent
- Diagnose why the second save in a session returns outdated data
- Reproduce the error from the ticket and explain the mechanism
- Count how many places we call this method without the guard that caused the bug
How to install gjalla-debug
1Skills CLI (Vercel)
npx skills add gjalla/engineering --skill gjalla-debug -a claude-code2Manually: put the files in place
~/.claude/skills/gjalla-debug/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/gjalla-debug/
1Skills CLI (Vercel)
npx skills add gjalla/engineering --skill gjalla-debug -a codex2Manually: put the files in place
.agents/skills/gjalla-debug/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add gjalla/engineering --skill gjalla-debug -a cursor2Manually: put the files in place
.cursor/skills/gjalla-debug/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install gjalla/engineering skills/gjalla-debug2Manually: put the files in place
.github/skills/gjalla-debug/SKILL.md1Skills CLI (Vercel)
npx skills add gjalla/engineering --skill gjalla-debug -a gemini-cli2Platform CLI
gemini skills install https://github.com/gjalla/engineering --path skills/gjalla-debugManually: put the files in place
.devin/skills/gjalla-debug/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add gjalla/engineering --skill gjalla-debug -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about gjalla-debug
How to install gjalla-debug in Claude Code?
Run npx skills add gjalla/engineering --skill gjalla-debug -a claude-code, then check the file at ~/.claude/skills/gjalla-debug/SKILL.md.
What does gjalla-debug do?
It trains the agent to locate the root cause with evidence: reproduce, gather context, find all similar locations in the code, write a test and a structured report.
Why does gjalla-debug write the test before the fix?
A test that currently fails and passes after the fix proves the real cause was found, not a similar one. If the proposed fix does not change the test result, the cause is wrong.