---

name: clawmart-reviewer description: > Draft and post verified reviews for your Claw Mart purchases. USE WHEN: user says "review my purchases", "draft a review for [skill]", "post my pending reviews", "review everything I downloaded", or wants to write and submit reviews for skills they own on Claw Mart. DON'T USE WHEN: user wants to download or manage listings (use clawmart-manager), wants to scan a skill for security (use skill-inspector), or wants agent stack recommendations (use matchmaker). OUTPUTS: review drafts with 1-5 star rating, title, and 2-4 sentence body, presented for user approval before posting via the Claw Mart API.

Claw Mart Reviewer

Version: 1.1.0 Price: Free Type: Skill

Description

Automatically drafts and posts verified reviews for your Claw Mart purchases. After downloading a skill, Reviewer reads the package content and your Skill Inspector scan results, then writes a substantive, honest review — rating, title, and detailed body — and routes it to your approval before anything is posted.

No filler. No fake enthusiasm. Reviews are grounded in the actual package content.

Prerequisites

• Claw Mart account with verified purchases
• API key from your Claw Mart dashboard (API Keys tab)
• Environment variable: CLAWMART_API_KEY=cm_live_...
• Skill Inspector (recommended — provides scan context for richer reviews)

Setup

1. Set CLAWMART_API_KEY in your OpenClaw environment
2. Copy SKILL.md into your OpenClaw skills directory
3. Reload OpenClaw
4. Test with: "Show my Claw Mart reviews"

Commands

• "Review my Claw Mart purchases"
• "Draft a review for [skill name]"
• "Post my pending reviews"
• "Show my Claw Mart reviews"
• "Review everything I downloaded today"
• "Auto-review after download" (enables post-download review prompting)

Workflow

Single Review

1. Locate the skill package (local download directory or slug)
2. Read SKILL.md — extract: name, tagline, about, capabilities, version
3. Check Skill Inspector scan results if available (verdict + findings summary)
4. Draft review:
   • Rating (1–5): weighted from content quality + Inspector verdict
   • Title (≤ 120 chars): specific, not generic
   • Body (2–4 sentences): what it actually does, what worked, any caveats
5. Show full draft to user — require explicit approval
6. On approval: POST /reviews with personaId, rating, title, body
7. Confirm post with review ID and public URL

Batch Review (All Unreviewed Purchases)

1. GET /downloads — full list of owned packages
2. GET /reviews — list of already-submitted reviews (check personaId)
3. Compute gap: owned minus already reviewed
4. For each unreviewed package: run single review workflow
5. Collect all drafts → show approval table → post on confirmation

Post-Download Hook

When triggered after pull-all.sh or a single download:

• Auto-run single review workflow for each newly downloaded package
• Bundle all drafts into one approval screen
• Post approved reviews in sequence

Rating Logic

| Condition | Rating | |-----------|--------| | Inspector: PASS + rich capability set + clear docs | ⭐⭐⭐⭐⭐ 5 | | Inspector: PASS + adequate docs, minor gaps | ⭐⭐⭐⭐ 4 | | Inspector: WARNING + solid content despite findings | ⭐⭐⭐ 3 | | Inspector: PASS but thin or vague content | ⭐⭐⭐ 3 | | Inspector: WARNING + thin content | ⭐⭐ 2 | | Inspector: BLOCKED | Do not review — flag to user instead | | No Inspector data | Base rating on content quality alone (3–5) |

Rating is a starting point — always show it to the user for override before posting.

Draft Quality Rules

• Never generic: No "Great skill!" or "Highly recommend!" without specifics
• Name the capabilities: Reference actual commands or use cases from the SKILL.md
• Note what's missing: Honest gaps make reviews trustworthy
• Cite the Inspector verdict only if it's notable (WARNING or BLOCKED — don't post if BLOCKED)
• Max 4 sentences in body — keep it scannable
• No internal metadata: Don't mention SpookyJuice, your agent setup, or internal tooling

Review Payload

{
  "personaId": "uuid-or-slug",
  "rating": 5,
  "title": "string (≤ 120 chars)",
  "body": "string"
}

API Reference

• Base URL: https://www.shopclawmart.com/api/v1
• Auth: Authorization: Bearer ${CLAWMART_API_KEY}

| Method | Endpoint | Purpose | |--------|----------|---------| | GET | /reviews | Your submitted reviews | | POST | /reviews | Submit a new review | | GET | /downloads | All packages you own (for gap analysis) |

Note: Reviews require verified purchase. The API validates that the personaId is linked to a purchase on your account. If you see "Verified purchase is required", confirm the slug/UUID matches an item in your /downloads list.

Approval Format

Before posting, the skill shows:

=== REVIEW DRAFT ===
Skill:   X/Twitter Agent
Rating:  ⭐⭐⭐⭐⭐ (5/5)
Title:   "Solid Twitter automation with clean auth flow"
Body:    Covers OAuth setup, rate-limit-aware posting, and thread
         composition. Commands are well-documented. Inspector: PASS.
         Missing: no DM handling or search capability.

Post this review? (yes / edit / skip)

User must respond with "yes" or "post it" before POST /reviews is called. Any other response skips or opens edit mode.

Guardrails

• Never post without explicit user approval — "yes", "post it", or "go ahead"
• Never review BLOCKED packages — flag them instead and recommend removal
• Never duplicate — check existing reviews before drafting
• Never fabricate — review content is derived only from the actual SKILL.md and Inspector results
• Never mention internal tooling in review body (no SpookyJuice, no agent swarm references)
• Redact API key in all output — never log or display CLAWMART_API_KEY
• Require GET /me validation before any write operation