skills

Diagnose software failures through reproduction, boundary localization, competing hypotheses, and discriminating experiments. Use when Codex is asked to investigate a bug, flaky test, crash, incorrect result, performance regression, production symptom, or unexplained behavior and should determine the cause before implementing a fix. Produce an evidence-backed causal explanation and verification plan; do not patch unless the user also asks for a fix.

Design, implement, or review compatible and reversible database, event, API payload, configuration, or persisted-data schema evolution. Use for migrations, column or field changes, backfills, constraint changes, data transformations, serialization upgrades, zero-downtime deploys, or rollback planning where old and new code or data may coexist. Require explicit expand-migrate-contract sequencing, data invariants, operational bounds, and recovery evidence.

Audit an AI agent harness for production readiness from repository code, configuration, tests, and run traces. Score 44 runtime controls across eight outcomes with artifact evidence, identify audit limitations, assign an evidenced maturity band, make a launch decision, and produce a dependency- ordered fix queue. Use for agent safety, governance, ship-readiness, control- gap, or due-diligence reviews of LangGraph, OpenAI Agents SDK, Google ADK, CrewAI, multi-agent, MCP-enabled, or custom agent systems. Do not use as a questionnaire or as a substitute for a runtime penetration test.

Produce an implementation-ready plan for a software change by inspecting the actual repository, tracing affected execution paths, resolving material design decisions, and defining verification and rollout evidence. Use when the user asks for an implementation plan, technical approach, task breakdown, design proposal grounded in an existing codebase, or wants to know how a feature or fix should be built before editing begins. Do not use for generic product roadmaps or when the user has asked to implement immediately.

Restructure existing code while preserving externally observable behavior through explicit invariants, characterization evidence, small reversible steps, and continuous verification. Use when the user asks to refactor, simplify, extract, reorganize, modularize, deduplicate, split a component, improve architecture, or pay down technical debt without intentionally changing product behavior. Distinguish refactoring from feature work and stop scope drift from hiding inside structural edits.

Map an unfamiliar software repository from code and configuration. Use when Codex needs to understand a new codebase, locate entry points or ownership boundaries, explain architecture and runtime flow, identify where a change belongs, or prepare for implementation without yet modifying files. Produce an evidence-backed repository map, not a directory listing or README paraphrase.

Review a code change for concrete correctness, security, data-integrity, compatibility, concurrency, and operability regressions by reconstructing intent and checking affected invariants and boundary paths. Use for local diffs, commits, branches, patches, or pull requests when the user wants actionable review findings. Report only defects introduced or exposed by the change with precise evidence; do not modify code unless asked to address findings.

Triage an active or recent software production incident by establishing impact, stabilizing the system, preserving evidence, building a timestamped timeline, coordinating hypotheses, and choosing reversible mitigations. Use when users report an outage, severe degradation, security or data-integrity event, broken deployment, elevated errors, or ask for incident command, status assessment, mitigation, or investigation support. Prioritize user harm reduction over root-cause completeness and require explicit authority for production mutations.

Design and execute a risk-based verification strategy for a software change, bug fix, refactor, migration, configuration update, or release candidate. Use when Codex needs to prove that a change works, choose proportional tests, investigate whether validation is sufficient, or report what remains unverified. Trace claims to evidence across static checks, focused tests, integration boundaries, and runtime behavior instead of merely running the entire test suite.

Write a clear, conventional git commit message from staged changes. Use when the user asks to commit, write a commit message, or describe what changed.

Draft a pull request title and description from the commits and diff on the current branch. Use when the user asks to open a PR, write a PR description, or summarize a branch's changes for review.

Improve a project's README for clarity, structure, and completeness without changing its meaning. Use when the user asks to clean up, polish, or improve a README or project documentation.

Replace with a description of what this skill does and when Claude should use it.