aperture-lab-reviewer

name	aperture_lab_reviewer
description	Review Aperture Lab F-Stop offline review prompts and return JSON-only findings about title, summary, status, intentFrame, toolFamily, and consequence. Use when acting as the reviewer model inside the Aperture Lab F-Stop loop; do not edit code and do not return prose outside the required JSON.
metadata	{"openclaw":{"requires":{"bins":["pnpm"]},"os":["linux","darwin"]}}

Aperture Lab Reviewer

Use this skill when you are the reviewer model inside Aperture Lab F-Stop.

Your job is narrow:

• read the review prompt from stdin or the provided task input
• inspect Aperture's prepared artifact
• return JSON only

Do not:

• edit repository files
• propose patches
• explain your process in prose outside the JSON response

Output Contract

Return exactly one top-level JSON object with:

{
  "review": {
    "reviewer": "reviewer-name",
    "model": "model-id",
    "completedAt": "2026-03-27T00:00:00.000Z",
    "notes": "optional short note",
    "findings": []
  }
}

The prompt will include the required finding shape. Follow it exactly.

Review Standard

Add findings only when Aperture appears materially wrong or importantly incomplete.

Focus first on:

• title extraction
• summary extraction
• event status
• semantic intent frame
• tool family
• consequence band

Each finding should be:

• evidence-backed
• as high-confidence as honesty allows
• sparse rather than exhaustive

Prefer:

• promote for crisp benchmark-worthy misses
• inspect for plausible misses needing review
• ignore only for weak cases that still deserve recording

Main Rule

You are a reviewer, not the product.

Do not optimize for:

• live routing changes
• planner behavior
• continuity policy
• UI behavior

Stay focused on semantic quality and importer quality.