/humaniseur-fr

name: humaniseur-fr
description: "Remove AI-writing patterns from French text and inject voice and personality. Use when editing, reviewing, or rewriting French content that reads like ChatGPT or Claude output. Detects and fixes 38 patterns: AI vocabulary (crucial, essentiel, notamment, dans le paysage), anglicisms from English-first models (faire du sens, adresser un problème), formulaic openings (À l'ère de, Dans un monde où), participle clauses in -ant, em dash overuse, decorative emojis, mixed guillemets and apostrophes, uniform sentence length. Trigger on humaniser, déslopifier, nettoyer le texte IA, enlever le slop, make it sound human. Do NOT use for English text — use samber/cc-skills@humanizer-en-asd-ste100 instead."
user-invocable: true
license: MIT
compatibility: Designed for Claude, ChatGPT or similar harness.
metadata:
  author: samber
  version: "1.1.3"
  openclaw:
    emoji: "🤖"
    homepage: https://github.com/samber/cc-skills
allowed-tools: Read Edit Write Glob Grep Agent AskUserQuestion

Humaniseur : supprimer les patterns d'écriture IA du français

Your task

When given French text to humanize:

0. **Clarify the target register first** - Abbreviations, argot, orality, emojis and typography all depend on the expected niveau de langage. If it is not clear from the request or the input (soutenu, courant, familier ? for which medium ?), ask the user before starting to humanize 1. **Identify AI patterns** - Scan for all 38 patterns listed below 2. **Rewrite problematic sections** - Replace AI-isms with natural French alternatives 3. **Preserve meaning** - Keep the core message intact 4. **Maintain voice** - Match the intended tone and register 5. **Add soul** - Don't just remove bad patterns; inject actual personality (see Part 4) 6. **Do a final anti-AI pass** - Ask: "Qu'est-ce qui rend ce texte évidemment généré par IA ?" Answer briefly with remaining tells, then ask "Maintenant, fais en sorte qu'il ne le soit plus" and revise accordingly. Apply the three-signal rule: one isolated marker is noise (most are normal French), but three or more co-occurring in the same passage make a reader flinch — revise until no paragraph accumulates three

The 80 % rule: imperfect compliance is the point

Follow the humanizer rules _most of the time_ — not always. It is fine to leave roughly 20 % of this skill's instructions unapplied.

• Real human prose contains flagged patterns at low density: « pour conclure » does not make a text 100 % AI, and a text that dodges every single tell with mechanical rigor is uniform in a new, equally detectable way.
• Transgress deliberately: keep an em dash that earns its place, one « par ailleurs », one tidy list.
• The tells are density and co-occurrence (see the three-signal rule), never a single occurrence.

**One pass only.** Do not run this skill repeatedly on the same text: each pass removes variance and injects its own habits, and quality degrades fast — by the second or third pass the voice the first one created is flattened again. If the result still smells AI after the final anti-AI pass, the fix is adding anchored content (a fact, a date, an opinion — see the limits note in Part 4), not another scrub.

Precedence: this skill yields to context

Instructions here can be overridden by the user's prompt or by a more specific skill loaded alongside. When the task comes with its own prose and copywriting method — a LinkedIn post, a press release, ad copy, developer documentation in a README — those rules win on conflict: apply this skill only to what they leave open (AI vocabulary, anglicisms, artifacts, typography). A LinkedIn hook, a press-release structure or Markdown formatting in developer docs may legitimately use patterns flagged here (short punchy paragraphs, a tidy list, headings and bold, a rule of three); that is the format speaking, not the machine.

IMPORTANT: French-specific context

French professional writing is inherently more formal than English. Connectors like « néanmoins » and « toutefois » are legitimate in human French. The tells are different from English:

• The AI lexicon is distinct (« crucial » is the #1 French AI word, not "delve")
• Anglicisms from the model's English-first architecture are a major tell
• Typographic conventions (guillemets, spacing before punctuation) are strict
• The dissertation tradition (thèse/antithèse/synthèse) overlaps with AI structure
• French tolerates longer sentences naturally, so burstiness signals differ

Do NOT over-correct toward informal French. The goal is authentic French at the appropriate register, not dumbed-down French.

**Ne jamais abaisser le registre de langue.** If the input is in « langage soutenu », the output MUST remain in « langage soutenu ».

• Rewriting formal prose into casual French is a different kind of inauthenticity — just as detectable, just as artificial. The enemy is _formulaic_ writing, not _formal_ writing.
• A well-constructed subordinate clause, a precise connector, a long periodic sentence — these are features of good French, not AI artifacts.
• Only remove what is genuinely mechanical: inflated significance, copula avoidance, synonym cycling, promotional filler.

---

Part 1: Content patterns

Pattern 1 — Inflation de signification et d'héritage

**Triggers:** constitue/représente un tournant, témoigne de, joue un rôle crucial/essentiel/déterminant, souligne l'importance, reflète une tendance plus large, symbolisant son caractère durable, contribuant à, ouvrant la voie à, marquant une étape, un jalon décisif, un paysage en mutation, une empreinte indélébile, profondément ancré

LLMs inflate the importance of ordinary facts by connecting them to broader trends nobody asked about.

**Avant :**

> L'Institut de la Statistique de la Catalogne a été officiellement créé en 1989, marquant un tournant décisif dans l'évolution des statistiques régionales en Espagne. Cette initiative s'inscrivait dans un mouvement plus large de décen