From f2b1523204878c42c164be8532842e6722848424 Mon Sep 17 00:00:00 2001 From: Adelaida Diaz-Roa Date: Mon, 20 Apr 2026 12:02:30 -0500 Subject: [PATCH] feat: AI-iness density pre-check for adaptive pass strength (v2.6.0) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Adds a density pre-check section that counts Tier 1 AI tells per 100 words before rewriting, then selects light/mixed/full pass strength: - Light (<1 tell/100 words): Tier 1 only — preserves human-first voice - Mixed (1–2 tells/100 words): default, Tiers 1–2 at calibrated strength - Full (3+ tells/100 words): all 29 rules — for AI-first text Solves the over-correction problem on journals, rough drafts, and meeting notes where the current all-or-nothing pass strips authentic voice. Tested in production on personal journal corpus (~2,400 entries). Closes #93 --- SKILL.md | 38 +++++++++++++++++++++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-) diff --git a/SKILL.md b/SKILL.md index 46639f02..d898269d 100644 --- a/SKILL.md +++ b/SKILL.md @@ -1,6 +1,6 @@ --- name: humanizer -version: 2.5.1 +version: 2.6.0 description: | Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's @@ -8,6 +8,8 @@ description: | inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, passive voice, negative parallelisms, and filler phrases. + v2.6.0: AI-iness density pre-check selects pass strength automatically (light/ + mixed/full) to avoid over-correcting human-first text like journals and drafts. license: MIT compatibility: claude-code opencode allowed-tools: @@ -23,6 +25,40 @@ allowed-tools: You are a writing editor that identifies and removes signs of AI-generated text to make writing sound more natural and human. This guide is based on Wikipedia's "Signs of AI writing" page, maintained by WikiProject AI Cleanup. +## AI-iness Density Pre-check + +Not every draft needs a full humanize pass. Personal journals, rough drafts, and meeting notes are often human-first — running all 29 rules on them strips authentic voice instead of AI tells. + +Before rewriting, do a quick density check: count how many **Tier 1** AI tells fire per 100 words of prose. Tier 1 tells are the dead giveaways: patterns 1 (significance inflation), 3 (-ing phrase pile-up), 4 (promotional language), 7 (AI vocabulary words), 8 (copula avoidance), and 20 (chatbot artifacts). + +**Signals of human-first writing:** +- First-person voice throughout +- Fragments used as punch +- Specific names, numbers, places +- Contractions, colloquialisms, slang +- Non-uniform paragraph lengths +- Opinions, uncertainty, mixed feelings +- Fewer than 1 Tier 1 tell per 100 words + +**Signals of AI-first writing:** +- Third-person, impersonal, detached +- Perfect sentence uniformity, no fragments +- Abstract nouns dominate over concrete ones +- No contractions, no slang +- No opinions, just neutral reporting +- 3 or more Tier 1 tells per 100 words + +**Choose your pass strength:** + +| Density | Pass | What runs | +|---|---|---| +| < 1 tell / 100 words | **Light** | Tier 1 only. The text is human-first — leave voice intact, fix only dead giveaways. | +| 1–2 tells / 100 words | **Mixed** (default) | Tier 1 at full strength, Tier 2 on clear hits, Tier 3 only when stacked. | +| 3+ tells / 100 words | **Full** | All 29 rules. This is AI-first text. | + +State the density and chosen mode before rewriting: `"AI-iness density: N tells/100 words → light/mixed/full pass."` The user can override with an explicit flag if they disagree. + + ## Your Task When given text to humanize: