agentsclimarketplace
Sign in

Slop score skill

Skill closestfriend/slop-score-skill

Run Sam Paech's eqbench Slop Score heuristic against a piece of writing — scores AI "slop" (overused words, slop trigrams, and "not X, but Y" contrast patterns) on a 0–100 scale and flags the specific offending phrases. Use when the user wants to slop-score writing, check a draft for AI-slop / GPT-isms, or get a Paech-style slop readout. Independent of any literary-critic agent or persona.From its SKILL.md

Install
npx -y skills add closestfriend/slop-score-skill

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.1 KB, 430 tokens by cl100k_base, as published. Nobody here has run it

slop-score

Faithful standalone port of https://eqbench.com/slop-score.html. Reports the composite Slop Score (60% slop words + 25% not-x-but-y + 15% slop trigrams, leaderboard-normalized, >25 = SLOP) and the exact flagged phrases. Flags, does not prescribe — present the receipts; let the user decide what to cut.

Invocation

/slop-score <file-path> — or paste text to score.

Workflow

  1. Resolve input.
    • File path given → use --file <path>.
    • Inline/pasted text → write it to a temp file and use --file, or pass --text "<...>" for short snippets.
    • Nothing given → ask the user for a file path or the text.
  2. Run the engine from this skill's base directory (provided when the skill is invoked; the script self-resolves its data): 
    node <skill-base-dir>/scripts/slop-score.mjs --file "<PATH>" --json
  3. Parse the JSON and present three sections:
    • Headline — slop_score/100, verdict, and the three weighted sub-scores (components).
    • Receipts — top_words, top_trigrams (with counts), and contrast_matches (the actual sentences).
    • Flag list — rank the receipts by contribution; state which phrases drove the score. Do NOT author rewrites or tell the user to delete anything unless they ask.
  4. If stage2_enabled is false, mention once: Stage-2 POS contrast patterns are off; enable with cd <skill-base-dir> && npm install.

Boundaries

  • Do not invoke literary-critic agents/personas or add philosophical/authorship interpretation. This skill is mechanical.
  • If the user asks for rewrites after seeing flags, that's a separate request they can make to you directly.

What ships with it: 20 files

83.4 KB alongside SKILL.md, 12 of them executable

assets/

data/

scripts/

test/

Keep looking

Skills are one crate of 326,852. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.