convez.ai
SkillEscrita & ConteúdoIntermediário
Triyansha

Slop Detector

Triyansha/AI-slop-detector

Detecta, pontua e corrige conteúdo AI-gerado com coaching educativo

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill Claude que detecta, avalia e corrige conteúdo gerado por IA de forma automática. Fornece um Score de Slop (0-100), identifica padrões típicos de geração automática, oferece corretivos específicos e coaching para evitar produzir textos genéricos ou clichês de IA.

Instalação

$npx convez-ai Triyansha/AI-slop-detector

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
2
Contribuidores
1

Casos de uso

  • Avaliar PRDs, emails e decks antes de enviar
  • Treinar equipes a reconhecer padrões de texto gerado por IA
  • Melhorar qualidade de conteúdo removendo automatismos linguísticos

Veredito

Ferramenta prática para quem precisa garantir autenticidade e qualidade em conteúdo escrito. O design modular (detectors + knowledge forkável) permite customização, mas requer compreensão clara de padrões de 'slop' para máxima efetividade.

README

A tradução para o português deste README ainda está na fila.

Slop Detector

A multimodal AI-slop detector, de-slopper, and coach — a pure-instruction Claude skill. Paste text (decks, design, audio, and data in later versions), get a Slop Score (0–100, lower = cleaner), the specific tells, a de-slopped fix, and coaching to stop generating slop.

Not an "AI-or-human" accuser. A scorer + coach: what reads as machine-made, how to fix it, and how to prevent it.

Install

npx skills add Triyansha/AI-slop-detector

This uses the Vercel skills CLI to copy the skill folder (skills/slop-detector/) into your Claude skills directory. Add -g to install globally:

npx skills add Triyansha/AI-slop-detector -g

Clone the repo and copy the skill folder into your Claude skills directory:

git clone https://github.com/Triyansha/AI-slop-detector.git
cp -r AI-slop-detector/skills/slop-detector ~/.claude/skills/slop-detector

Then restart your Claude session so the skill is discovered.

Use

In Claude, with the skill installed:

  • "Is this AI slop?" / "Score this PRD" → detect
  • "Clean this up" / "de-slop this email" → de-slop
  • "Why does this sound like AI?" → coach

How it works

Router (SKILL.md) → five detectors → shared rubric + knowledge → de-slop / coach actions. The taxonomy is data (knowledge/), so it's forkable and gets smarter over time.

The taxonomy gets smarter over time: actions/update-taxonomy.md proposes newly-emerging tells as dated candidate digests in trends/, and npm run promote appends reviewed ones (gated by the test suite).

Repo layout

  • skills/slop-detector/ — the installable skill (what npx skills add copies): SKILL.md, detectors/, actions/, knowledge/, rubric.md, interrogator.md, trends/.
  • tests/, tools/, package.json — maintainer dev harness (the zero-dependency validators + the promote helper). Not needed to use the skill; used to develop and grow it.

Develop / test

npm test          # zero-dependency contract + rulepack + regression-gate validators
npm run eval      # optional: scores the corpus via the Anthropic API (needs ANTHROPIC_API_KEY)
npm run promote -- trends/<date>.candidates.json   # dry-run promote of reviewed tells (add --apply to write)

Status

v1.0.0 — Five detectors + intelligence loop, installable via npx skills. The full foundation is live.

License

MIT

Ver o original no GitHub

Bom para

Relacionados