Recipe: Code Review

On-demand entry point for the plan-marshall code-review capability. This recipe runs a focused, diff-aware structural and quality review over the current footprint and emits each discovered issue as a lint-issue finding into the triage pipeline — the structural difference from external review tools, which print a report and stop.

The recipe is standalone: it does NOT load the shared security-audit engine (recipe-security-audit/standards/audit-engine.md). It reuses only the diff-aware shape that family shares — compute the live footprint, review it, emit findings, dispatch to triage — applied through the code-reviewer lens rather than the security one. The review itself is an LLM cognitive review, consistent with the recipe-skill design model (recipe skills are mode: workflow cognitive procedures, not deterministic scripts).

The reviewer reads the footprint through the plan-marshall:persona-code-reviewer lens — the meta/evaluator persona that composes the work personas (implementer, module-tester, integration-tester, documenter, security-expert) as evaluation lenses, judging correctness and intent rather than authorship.

Foundational Practices

Skill: plan-marshall:persona-plan-marshall-agent

Enforcement

Execution mode: Walk the four steps below in order over the footprint — compute footprint, gather the reviewer lens, run the cognitive review, emit findings to triage. Each step has a single explicit job — no improvisation, no extra discovery passes.