claude-skills

Cross-validate Claude Code's implementation plan or code changes using OpenAI Codex CLI as an independent second opinion. Runs an iterative review loop: Codex reviews, Claude dispatches subagents to evaluate each feedback point in parallel (using feedback-review analysis), implements accepted changes via parallel subagents, then sends a summary back to the same Codex conversation for re-review. Repeats until Codex approves or has no further actionable feedback. Triggers on phrases like "codex review", "double check with codex", "second opinion", "cross-validate", or "verify with codex".

Analyze feedback on code from any source — teammate PR reviews, AI-generated flags, implementation review findings, or pasted critique — and vet each point for correctness, optimality, and elegance before acting on it. Uses parallel subagents for thorough investigation. Triggers on phrases like "got this feedback", "teammate review", "review comments", "PR feedback", "address these findings", "handle this review", or when the user pastes numbered/bulleted critique, flag output, or structured review results.

Triage and act on AI-generated code review flags. Use this skill whenever the user pastes a list of review flags, code review findings, AI reviewer output, or automated code analysis results that need to be investigated and resolved. Also trigger when the user asks to "review flags", "triage findings", "check these issues", or shares a batch of code concerns with file paths and descriptions — even if they don't call them "flags" explicitly.

Generate a structured recap of an implementation — the original problem, the solution, and a defense of why it's the right approach across correctness, optimality, and elegance. Use this skill when the user asks to "recap", "summarize the implementation", "write up the changes", "explain what we did and why", "document the approach", or wants to articulate the rationale behind a set of changes for a PR description, design doc, team update, or their own understanding. Also trigger when the user wants to explain or justify an implementation to others.

Thorough review of code changes for correctness, optimality, and elegance using parallel subagents. Use this skill whenever the user asks to review an implementation, audit recent changes, review a diff or PR, check code quality, or wants a second opinion on changes they or someone else made. Also trigger when the user says things like "review this", "how does this look", "check my changes", "audit this code", or "review the implementation" — even if they don't specify the axes explicitly.