review-areas
Area-focused code review that fans out parallel subagents — each examining a different review perspective (correctness, tests, security, performance, product) — then synthesizes only high-signal findings.
Skills
| Skill | Description |
|---|---|
| review-areas | Fan out area-specific subagents in parallel and merge findings by severity |
How It Works
- Scope — pick up the active PR, git diff, or explicit file list.
- Fan out — launch 2–4 parallel subagents, each with a focused area prompt (correctness, tests, security, performance, product). No custom agents needed — each subagent gets a self-contained prompt.
- Synthesize — deduplicate across areas, order by severity, apply a strict signal filter.
- Fix or report — by default, fix well-defined issues immediately; stop at reporting if the user asks for review-only.
Comparison with model-council
| review-areas | model-council | |
|---|---|---|
| Diversity source | Different areas (correctness, security, …) | Different models (GPT, Opus, Gemini) |
| Custom agents | None — uses unnamed subagents | Three model-pinned agents |
| Best for | Thorough single-model deep dives | Cross-model consensus and blind-spot detection |
Both plugins can be used together: run review-areas for area depth, then model-council for model diversity.
Plugin Structure
review-areas/
├── plugin.json
├── README.md
└── skills/
└── review-areas/
└── SKILL.md