ad-review

Automated code review against governance standards is done by reading project diffs and comparing them to core documents like AGENTS.md, ARCHITECTURE.md, and GUIDELINES.md. This produces separate Standards Findings and Spec Findings in a single session.

Adversarial code review is a single-pass process that audits software diffs against binding project context and architecture rules. It generates formal in-session findings and creates an audit trail for traceability without spawning subagents.

Generating a code review audit trail is achieved by writing review results directly to the .agentic/reviews/ directory. This ensures traceability and provides a record for future re-runs of the software change assessment.

The code review workflow explicitly avoids spawning subagents during execution. It operates entirely in a single session to assess diffs and produce a formal handoff along with in-session findings.

Architecture decision records (ADRs) are read alongside CONTEXT.md and other governance documents to ground code change assessments. This binding context ensures the review evaluates diffs against established architectural decisions.

A formal handoff for code review findings is needed when completing a single-session adversarial audit of software diffs. It finalizes the review by summarizing Standards Findings and Spec Findings for downstream traceability.