changes-review

Codex compatibility note:

• Invoke $skill-name: load the registered .agents/skills/<name>/SKILL.md and execute it with Codex tools; no separate Skill tool or Claude mode switch is needed. When a workflow step or another skill names a skill, read that file in full yourself (every line, with no line limit; in parts only when output is cut off): a skill whose implicit invocation is off is never loaded for you. Author .claude/**; mirrors are generated.
• Track steps before execution and keep progress/evidence current. Explicit skill requests execute their protocol. Use the native question tool for required questions.
• Map foreign tool names to available host capabilities. A name mismatch needs no approval; an actually unavailable required capability stops with its step/evidence and a user question before adaptation.
• Subagent authorization: when a skill is user-invoked or AI-detected and its protocol requires subagents, that skill activation authorizes use of the required spawn_agent subagent(s) for that task.
• Preserve protocol order unless the user approves a deviation. Workflows use $start-workflow: gates stay fixed; other steps may flex with logged reasons and evidence.

Quick Summary

Goal: Find real defects in the requested diff and, when requested, converge its fixes without losing coverage or quality.

Summary: Triage every changed file and write a short plan → create review tasks, including domain entity quality and performance-test quality when those surfaces change → check connected behavior and applicable concerns → validate findings → report once or fix and freshly re-review. Choose review depth and delegation from risk and available context.

Workflow: Resolve scope → plan/tasks → review, with the whole-target $why-review rationale pass and a $performance-review pass when the change can affect performance → validate → review-only report or fix/re-review → final checks and recap.

Key Rules: Always invoke $why-review on the whole target alongside applicable specialist reviews, even when there are no findings. Review-only is the standalone default; --fix-loop enables the shared three-round loop. --loop-owner=caller returns a read-only pass to its fixing coordinator. Use the carried review-policy protocol for completion and user-approved extensions.

Todo first: create tasks for triage, review, validation, fixes, fresh re-review and final checks before execution; keep one task in progress and attach evidence when completed.

Contents