Code Review gives your agent a skeptical second pair of eyes on any change. It does not confirm your code looks fine. It assumes there is a bug. It tries to prove the bug exists. It catches real problems that a rubber-stamp review misses.

You point it at a diff or a PR. It runs an adversarial review. A reviewer reads the change. It assumes the change contains a bug. When Gemini is available, a second reviewer does the same from a different model. The two reviewers do not share one model's blind spots. Each reviewer returns an actionable result. It returns a failing test, a DEFECT: file:line, or NO ISSUES FOUND. It never returns loose prose.

When you install development-lifecycle, these reviewers run automatically at the /review phase. An evidence-tracker hook stops a task from being ticked off. The agent must read the spec and test files first.

Direction	Plugin	How
Enhances (soft)	Development Lifecycle	Runs the adversarial reviewers at /review and guards /work's checkbox flips — only when both are installed.
Enhanced by (soft)	—	None.
Requires (hard)	—	None. Code Review is fully standalone.
Required by (hard)	—	None.

Code Review is opinionated. It does not fit every workflow. Choose something else if:

• You trust another review pass. This pass could be a human reviewer, a linter suite, or a different AI reviewer. You do not want a second opinion on every change.
• You want a reviewer that explains its reasoning. This reviewer deliberately answers only with a failing test, a DEFECT: file:line, or NO ISSUES FOUND.
• You want the reviewer to assume the code is fine. The assume-a-bug framing is adversarial. It feels like overkill on small or throwaway changes.

Each primitive links to the source that implements it.

Primitive	Kind	What it does
/code-review	command	Adversarial review of a diff or PR.
/doubt	command	Fresh-context review of a decision before it stands.
/simplify	command	Cut accidental complexity from a change.
security-review	skill	Three-tier boundary security review.
testing-strategy	skill	Coverage audit — the Beyoncé Rule, DAMP, Prove-It.
adversarial-reviewer	sub-agent	Assume-bugs critic; returns a failing test or a DEFECT.
adversarial-reviewer-cross	sub-agent	The same, cross-model via Gemini; degrades visibly — a CROSS-REVIEW-DEGRADED: <reason> line, never a silent same-model swap, whenever cross-review.sh exits non-zero: agy missing or failing, its print timeout running out, material over the size ceiling, or a reply that breaks the output contract twice.
security-auditor	sub-agent	Finds the unvalidated boundary crossing.
test-engineer	sub-agent	Finds the behavior with no test.
evidence-tracker	hook	Blocks a [ ] → [x] flip until specs and tests are read (Claude-only).

You do not need to configure anything. The plugin works out of the box.

• First code review — the tutorial.
• Review a change · Simplify a diff · In-flight decision review — the how-tos.
• Why adversarial review — why the assume-bugs framing works.
• Code-review design · Composition design — the deeper design.

Reference · Architecture · Home