ClawHub

Slop Cop

Other

Judges visual design assets and AI-generated images before they ship. Use when the user wants to compare design options, choose between asset variants for a hero/banner/icon slot, vet AI-generated images for hallucinated text or artifacts (extra fingers, gibberish, melted hands, watermarks), audit a landing page's visual choices, or get a second opinion on visual design before deploying. Triggers on phrases like "which is best as a hero", "is this image safe to ship", "does this fit the brand", "pick between these assets", "audit this banner", "check this for AI slop", or general "is this slop?" requests.

Install

openclaw skills install slop-cop

slop-cop

A visual-design referee. Given one or more image assets plus a decision context, produce strict per-asset verdicts (SHIP, FIX, or KILL) and, when multiple candidates compete for one slot, a ranked recommendation with placement reasoning.

The goal: stop hallucinated text, melted hands, off-brand vibes, and obvious AI artifacts from reaching production.

When to invoke

  • User has 1–N images and a decision to make ("which works best for hero?", "is this safe to ship?", "does this fit my brand?").
  • User wants a second opinion on a visual choice before deploy.
  • User asks to audit a landing page or compare AI-generated variants.
  • User explicitly says "slop check" / "is this AI slop?"

Inputs the skill needs

Before analysis, confirm or infer:

  1. Image paths — 1 or more local file paths or URLs.
  2. Decision context — what slot/role is this for? Examples: "hero banner at 1200x600", "square avatar 1024x1024", "mobile card at 4:5", "is this safe to ship anywhere?".
  3. Target render size / aspect ratio — if relevant (e.g. hero rendered at 600x600 with rounded corners and a 4px border).
  4. Brand palette / style — hex colors and a one-line style descriptor when available (e.g. "navy #0f3a66 / orange #f3812a, cartoon illustration").
  5. Mode — single-asset audit (SHIP/FIX/KILL) or comparative pick (rank + recommend one).

If the user does not provide brand context, ask once. If they decline, proceed without brand-fit scoring and note it in the verdict.

Workflow

1. Run the vision pass

For each image, call the OpenClaw image tool with the strict checklist prompt in references/vision-prompt-template.md. Pass one image per call when possible — keeps the model focused. Use images (multi) only for explicit side-by-side comparison once each has been individually vetted.

The prompt template forces the vision model to enumerate findings against a fixed checklist instead of writing vibes-based prose.

2. Score against the full checklist

The mandatory checklist lives in references/checklist.md. Every asset must be scored on:

  • Hallucination scan — gibberish text, extra/melted fingers, broken anatomy, duplicated objects, watermarks, AI signatures, lighting that contradicts itself.
  • Legibility at target size — can any text on the asset be read at its actual render size?
  • Responsive safety — will the focal subject survive cropping to 16:9, 4:5, 1:1, and 9:16? Identify the focal point in pixel/percent terms.
  • Cross-browser / format — transparency needs (PNG/WEBP vs JPG), color profile concerns (sRGB vs P3), iOS Safari quirks.
  • Brand fit — if palette/style provided, check coherence; flag major mismatches.
  • Format / size sanity — actual dimensions, file size for web, aspect-ratio fit for the target slot.

3. Assign a verdict per asset

Use exactly one verdict word per asset, plus a one-sentence reason. No hedging, no "looks okay but...".

VerdictMeaning
SHIPClean. Deploy as-is.
FIXSalvageable with a specific edit (crop, recolor, regenerate text region, swap to different aspect). State the fix.
KILLDo not use. Hallucination, off-brand, broken anatomy, or wrong-tool-for-the-job.

Hard kill triggers (any one of these = automatic KILL):

  • Visible hallucinated/gibberish text on a graphic shipping to prod.
  • Extra/missing/melted fingers on a human or human-adjacent character.
  • Visible watermark or AI-tool signature.
  • Major brand-palette violation (when palette provided) that can't be fixed by recolor.

See references/anti-patterns.md for the full kill list and CSS-level gotchas that come up on real sites.

4. Comparative mode (multiple candidates, one slot)

When the user is choosing between assets for a single slot:

  1. Verdict each candidate individually first.
  2. Drop all KILL verdicts from the running.
  3. Rank remaining SHIP and FIX candidates by fit-to-context (brand match > focal-point survival > legibility > polish).
  4. Recommend one. Name the file path, the slot, and one-sentence placement reasoning.
  5. If every candidate is KILL or FIX, recommend regeneration with a brief brief.

5. Output format

Return a structured response:

## Verdicts
- <filename> — VERDICT — one-sentence reason
- <filename> — VERDICT — one-sentence reason
...

## Anti-patterns flagged
- (optional) bullet list of CSS/HTML/format gotchas detected from context

## Recommendation
<For comparative mode: which file goes in which slot, why, and any FIX steps needed before deploy.>

## Deploy notes
<Concrete file paths, target dimensions, format conversions, and any CSS/HTML lines that should change. Do NOT execute deploys — describe them.>

Keep it tight. No filler, no "great question."

Failure modes

  • Vision tool unavailable / errors out — Document the failure, then make a best-effort judgment from filename, file metadata (identify / file / exiftool if available), and decision context. Mark the verdict as BEST-EFFORT in parens and flag that a manual eyeball is required before ship.
  • No brand context provided — Proceed; note "no brand check performed."
  • Asset is a wordmark/logo — Skip hallucination scan for stylized typography (intentional design ≠ gibberish), but still check legibility, format, and brand-palette match.

References