This skill gives a coding agent a diagnose-first process for bug reports, regressions, runtime errors, failing builds and broken tests. It reproduces the problem as narrowly as it can, tests hypotheses that can be proven wrong, finds the shared root cause and checks the smallest change that would fix it. It only applies a fix when you authorize one; otherwise it stops at the diagnosis and a structured report. You get a reproduced bug, its root cause and a proposed fix you choose whether to apply.
Investigate (gstack)Investigates bugs and product issues with a structured evidence-gathering loop before proposing fixes.Coding agent
Systematic debugging (Superpowers)Applies a disciplined debugging process for bugs, test failures, and unexpected behavior before proposing fixes.Coding agent
Debug a problem step by step, from repro to fixAnthropic's debug skill, from its engineering plugin, walks through a problem from reproducing it to isolating, diagnosing and fixing it. You start it with an error message, a stack trace or a description such as a bug that appeared after a deploy or only in production. It returns a debug report covering how to reproduce the issue, the root cause, the fix and how to stop it coming back. With connectors it can also pull in logs and recent deploys, related commits and existing bug tickets.Coding agentAnthropicKnowledge work plugins
Check GitHub Actions workflows that run AI agents for injection risksTrail of Bits' agentic-actions-auditor skill reviews GitHub Actions workflows that call AI coding agents such as Claude Code Action, Codex, Gemini CLI or GitHub's AI Inference action. It finds the steps that start an agent, traces into shared actions and called workflows, and records how each is configured. It then reports places where text an outsider controls, like an issue title, can reach the agent, along with loose sandbox settings and allowlists that accept any user. It only reads workflow files and does not run them.Coding agentSecurityCI/CD
Get a private repository ready to publish as open sourceTrail of Bits' open-sourcing skill takes a repository through the checks needed before it goes public. It starts with a hunt for secrets in the code and its history, because that decides whether the repository can be kept or must be recreated, then covers the license, documentation, tests and CI, repository settings, release versioning and language-specific packaging. The goal is that a newcomer can build and contribute to the project and nothing private leaks. Some steps change repository settings, so review them before you apply them.Coding agentOpen sourceRelease
Make small code changes in lazy senior dev modeThis skill puts a coding agent in a lazy senior developer mode that makes the smallest possible change and still finishes every part of the task. Before writing code it lists everywhere the change must reach, including callers, tests, fixtures, and config, and checks what could break for users. It prefers existing code, the standard library, or installed dependencies over new code, and ends each reply with what it skipped and any risk. It has three levels, lite, full, and ultra, switched with a slash command.Coding agentCoding
