claude-code-skills

Validation gate for freshly written code — run two independent reviews (codex CLI + a Claude review subagent) over the same scoped diff, then empirically vet every finding against ground truth before accepting or vetoing it. Use after landing a work item (especially subagent-written code) and before committing.

Work through a queue of Technical Project Plans serially — delegate each to a TDD subagent, double-review the result with codex and a Claude /review subagent, empirically vet every finding before accepting or vetoing it, and land one coherent commit per plan. Use when executing a roadmap of plan files (e.g. a _todo/ directory) for a port, migration, or multi-stage feature.

Review code for potential issues and improvements. Use when asked to review specific files, functions, or code sections.

Update the active Technical Project Plan for handoff when context is running low or the session is ending, so the next session continues instead of restarting.

Work on a Technical Project Plan — read the plan, identify the current phase, do that phase's work, and update the plan with discoveries. Use when starting or resuming multi-session work tracked in a plan file.

Plan and execute coherent Conventional Commit groupings for large, tangled working tree changes across many files.

Iterative deep planning with critiques and alternatives. Use when facing complex design decisions requiring thorough analysis.

Review the git staged diff for verified bugs before committing, then drive a clean Conventional Commit.

Use when committing work from the current session to stage ONLY hunks the session touched, not the entire file. Prevents accidentally staging unrelated uncommitted changes from other work.