from specialist-agent12
A four-phase, evidence-first debugging methodology agent skill: gather evidence, analyze patterns, formulate hypotheses, and implement & prove fixes.
Adds a structured debugging workflow for agents and developers. The skill prescribes a four-phase method (Gather, Analyze, Formulate, Implement) with clear gating outputs for each phase, templates for evidence and analysis, reproducible test design, and a three-strike rule to prevent blind fixes. It focuses on reproducibility, hypothesis-driven testing, and creating reversible checkpoints.
Invoke when encountering bugs, test failures, or unexpected behavior—especially when previous quick fixes failed. Use it before proposing or applying fixes to ensure root-cause understanding, to design minimal tests that reproduce problems, and to produce a verifiable fix + regression test. Useful for agents assisting in code review, incident triage, or developer workflows.
Designed for agents with Read/Edit/Bash capabilities (Copilot, Codex, Claude Code, Cursor) that can run diagnostic commands and author patches or commit messages.
This skill has not been reviewed by our automated audit pipeline yet.
Test-Driven Development (TDD)
Run disciplined TDD cycles: write a failing test, implement the minimal fix, then refactor — for features and bugfixes where correctness matters.
/verify — Verification Before Completion
Run fresh verification commands (tests, build, lint, types) and require full output evidence before claiming work is complete or merging changes.
/learn — Interactive Learning Mode
Teaching-first implementation assistant: explains why decisions are made, walks through architecture, and produces incremental code with pedagogy for onboarding
Adaptive Planning (/plan)
Generate an implementation plan matched to task complexity — from quick mini-plans to full design documents — before writing code or making large changes.
Grill (Adversarial Code Challenge)
Performs adversarial stress-testing of code to find edge cases, race conditions, security holes, and logical flaws before deployment.
Smart Commit (conventional commit generator)
Generate and apply conventional commit messages with automatic type & scope detection plus pre-commit validation for secrets and debug artifacts.
Conversion Rate Optimization (CRO)
Audit landing pages and user flows to find conversion friction and deliver prioritized, testable A/B recommendations.
