adversarial-review

SKILL.md

Adversarial code review using opposite-model reviewers to challenge work from distinct critical lenses.

Spawns 1–3 reviewers (Skeptic, Architect, Minimalist) on the opposing model's CLI to avoid same-model bias and ensure genuine adversarial critique

Reviewers attack based on brain principles and assigned lenses; produces a synthesized verdict without auto-applying changes

Scales reviewer count by change size: 1 reviewer for <50 lines, 2 for 50–200 lines, 3 for 200+ lines or 5+ files

Requires explicit intent statement before review and applies lead judgment to filter false positives and style-over-substance findings

Adversarial Review

Spawn reviewers on the opposite model to challenge work. Reviewers attack from distinct
lenses grounded in brain principles. The deliverable is a synthesized verdict — do NOT make
changes.

Hard constraint: Reviewers MUST run via the opposite model's CLI (codex exec or
claude -p). Do NOT use subagents, the Agent tool, or any internal delegation mechanism as
reviewers — those run on your own model, which defeats the purpose.

Step 1 — Load Principles

Read brain/principles.md. Follow every [[wikilink]] and read each linked principle file.
These govern reviewer judgments.

Step 2 — Determine Scope and Intent

Identify what to review from context (recent diffs, referenced plans, user message).