cross-model-review

# cross-model-review

## Metadata
```yaml
name: cross-model-review
version: 2.0.0
description: >
  Adversarial plan review using two different AI models.
  v2: Alternating mode — models swap writer/reviewer each round.
  Fully autonomous loop — no human input between rounds.
  Use when: building features touching auth/payments/data models,
  plans that will take >1hr to implement.
  NOT for: simple one-file fixes, research tasks, quick scripts.
triggers:
  - "review this plan"
  - "cross review"
  - "challenge this"
  - "is this plan solid?"
  - "adversarial review"
```

---

## When to Activate
Activate this skill when the user:
- Says any trigger phrase above
- Shares a plan and asks for adversarial/second-opinion review
- Asks you to "sanity check" a multi-step implementation plan

Do NOT activate for: simple fixes, one-liners, pure research tasks.

---

## Modes

### Static Mode (v1 — backward compatible)
Fixed roles: planner always writes, reviewer always reviews. Requires human to trigger each round.

### Alternating Mode (v2 — recommended)
Models swap roles each round. Fully autonomous — no human input between rounds.

**Flow:**
- Round 1: Model A writes the plan. Model B reviews.
- Round 2: Model B rewrites (based on its own review). Model A reviews.
- Round 3: Model A rewrites (based on its own review). Model B reviews.
- ...continues alternating until both agree (reviewer says APPROVED) or max rounds hit.

**Why this works:**
- Each model must implement its own critique — can't nitpick without owning the fix
- The other model catches over-engineering or proportionality issues
- Natural convergence: each round addresses the other's concerns

---

## Autonomous Orchestration (Alternating Mode)

You (the main agent) run this loop. It's fully autonomous after kickoff.

### Step 1 — Save the plan and init

```bash
node review.js init \
  --plan /path/to/plan.md \
  --mode alternating \
  --model-a "anthropic/claude-opus-4-6" \
  --model-b "openai-codex/gpt-5.3-...

# cross-model-review

![CI](https://github.com/Don-GBot/adversarial-review/actions/workflows/ci.yml/badge.svg)

Runs implementation plans through an adversarial review loop between two different AI models — a reviewer and a planner — iterating until the reviewer approves or maximum rounds are hit. The reviewer challenges the plan, the planner (you, the main agent) revises it, and this continues until all CRITICAL and HIGH issues are resolved. Think of it as a built-in second opinion that can't be gamed because it's a different model from a different provider.

---

## When NOT to Use

Skip this skill for:
- Simple one-file fixes or quick scripts
- Pure research or investigation tasks
- Changes you can fully reverse in under 5 minutes
- Plans already reviewed by a human engineer in the last hour

Using it on trivial tasks adds overhead without benefit.

---

## Prerequisites

- **Node.js >= 18.0.0** (uses `structuredClone`, `fs.readSync` on fd 0, etc.)
- **OpenClaw** — the skill is invoked by the OpenClaw agent and relies on `sessions_spawn` for the reviewer
- No npm install needed — zero external dependencies

---

## Quick Start

**Install from ClawHub:**
```
/install cross-model-review
```

**Trigger phrases** (say any of these with a plan in context):
- "review this plan"
- "cross review"
- "challenge this"
- "is this plan solid?"

The skill activates automatically and manages the full loop. No manual setup needed.

---

## Complete CLI Transcript (end-to-end)

This shows what the agent runs under the hood for a 2-round review that ends in approval.

```
# Step 1: write plan to temp file
$ cat /tmp/review-1234/plan.md
# Auth Redesign Plan
...

# Step 2: initialize workspace
$ node scripts/review.js init \
    --plan /tmp/review-1234/plan.md \
    --reviewer-model openai/gpt-4 \
    --planner-model anthropic/claude-sonnet-4-6 \
    --out tasks/reviews
tasks/reviews/2026-02-21T15-00-00-a1b2c3d4

# Step 3: round 1 — agent spawns reviewer, saves response, parses it
$...