
from claude24
A structured root-cause debugging workflow: investigate, understand, reason, and fix — with commands and templates for code search, GitHub issue lookup, log ana
Debug codifies a disciplined root-cause analysis workflow for software issues. It guides an investigator through evidence collection (code search, logs, GH issues), systematic reasoning (first principles, differential diagnosis), and fix generation (specific code changes and verification steps). The skill provides command templates for rg, gh, git, log analysis, and prompts for AI assistants to synthesise findings.
Use Debug when faced with reproducible bugs, flaky tests, performance regressions, or errors with unclear origins. It's especially valuable in complex codebases where multi-source evidence (code, history, issues, logs) must be combined to find THE root cause rather than surface symptoms.
Designed for agents and engineers that can run shell commands, query GitHub, and call large-model assistants for reasoning. Works well with developer-assistant agents (Gemini/GPT-style) and CLI-integrated automation.
A structured root-cause debugging workflow skill that enforces Investigate→Understand→Reason→Fix protocol. Contains no scripts — purely a reference document of bash commands using rg, gh, and gemini CLI. The root cause protocol and forbidden shortcuts table are valuable, but the skill is essentially a long cheat sheet of shell one-liners with heavy reliance on the Gemini CLI for reasoning steps.
Pure documentation skill with no scripts. Monolithic single-file structure with no scripts/ or references/ directories. Heavy coupling to Gemini CLI — most reasoning steps are gemini invocations rather than agent-native reasoning. The 'Forbidden Shortcuts' table is a genuinely useful pattern worth highlighting.