{"slug": "blader-humanizer-wikipedia-pattern-skill-to-strip-ai-writing-tells", "title": "blader/humanizer: Wikipedia Pattern Skill to Strip AI Writing Tells", "summary": "The blader/humanizer agent skill trended on GitHub in week 45 of 2026 with about 66.2k stars as of November 8, 2026, rewriting prose against the 26 patterns in Wikipedia's Signs of AI writing list, according to the repository's README. The MIT-licensed skill installs in Codex via `npx skills add blader/humanizer --global --agent codex` and triggers in Claude Code with `/humanizer:humanizer`, running a pattern list, rewrite pass, critique pass and final pass described in its open SKILL.md. The README states detector evasion is explicitly not a goal and cites a 16/16 judge preference in issue #229 as a community report rather than a peer-reviewed benchmark.", "body_md": "The [blader/humanizer](https://github.com/blader/humanizer?ref=explainx) repository trended on GitHub in **week 45 of 2026** and shows **about 66.2k stars** on its GitHub page as of November 8, 2026. It is not a model checkpoint. It is an **[agent skill](https://github.com/blader/humanizer/blob/main/SKILL.md?ref=explainx)** that rewrites prose using **[Wikipedia’s Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing?ref=explainx)** — the same pattern list WikiProject AI Cleanup maintains for encyclopedia editors.\n\nIf you ship docs, changelogs, or blog drafts through **Claude Code** or **Codex**, Humanizer is a concrete answer to a question explainx.ai keeps hearing: **how do I keep agent output from sounding like default ChatGPT cadence** without lying to readers?\n\n## TL;DR\n\n| Question | Answer | \n|---|---|\n| License | **MIT** (README) | \n| Primary source for patterns | **[Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing?ref=explainx)** | \n| Pattern count | **26** in README tables | \n| Claude Code trigger | **`/humanizer:humanizer`** (plugin path in README) | \n| Codex install | **`npx skills add blader/humanizer --global --agent codex`** | \n| Detector evasion? | **Explicitly not a goal** (README) | \n| Blind test claim | README cites **16/16** judge preference in issue**#229** — community report, not a peer-reviewed benchmark | \n\n## Why GitHub is amplifying it now\n\nTrending skills usually ride a **felt pain**. explainx.ai’s [Kevin Boone piece on trusting AI web content](https://explainx.ai/blog/kevin-boone-ai-generated-web-content-trust-gutenberg-2026) documents how readers downgrade pages that read like uniform model voice. Humanizer attacks **style tells**, not factual hallucination — complementary to [harness false-alarm research](https://explainx.ai/blog/andy-hall-ai-audit-agents-false-alarms-malt-r-judge-2026) on audits.\n\nThe repo README says it is **sponsored by cfo.ai**. Sponsorship does not change the skill mechanics: pattern list, rewrite pass, critique pass, final pass — all described in open **`SKILL.md`**.\n\nCompared with [benjitaylor/skills Stageshow](https://explainx.ai/blog/benjitaylor-skills-stageshow-faux-product-ui-agent-skill-react-github-trending-guide-2026) or [claude-animated-shorts](https://explainx.ai/blog/claude-animated-shorts-remotion-threejs-agent-skill-github-trending-guide-2026), Humanizer is **text-only** and **editorial**. It fits next to [jakubkrehel skills for UI](https://explainx.ai/blog/jakubkrehel-skills-better-interface-14-agent-skills-ui-typography-accessibility-github-trending-guide-2026) in a growing **skills marketplace** pattern explainx.ai tracks in [what are agent skills](https://explainx.ai/blog/what-are-agent-skills-complete-guide).\n\n## The before/after the README proves\n\nThe README’s launch example is a **shared drafts** feature announcement. Before text stacks **Not X but Y**, emoji rockets, **“Let that sink in.”**, and inflated significance. After text states **what shipped**, **what it cost in files**, and **pricing** in plain declarative sentences.\n\nThat example is marketing prose, but the same patterns show up in **internal RFCs**, **support macros**, and **agent-generated PR descriptions**. Humanizer is aimed at those defaults.\n\n## The five patterns it edits on sight\n\nRounded page with a pen and green spark, matching prose cleanup in agent workflows\n\nThe README ranks patterns by strength. The top five:\n\n1. **Not X but Y** — “It is not just a feature, it is a shift.”\n2. **One-line closers** — dramatic fragments after every section.\n3. **Sayings that sound deep** — faux philosophy instead of claims.\n4. **Staged run-ups** — “Here is the thing” before the fact.\n5. **Arguing with nobody** — straw objections nobody raised.\n\nPatterns marked **weak alone** need **clusters** before a rewrite — the README warns careful human writers may use any one deliberately.\n\n## Full pattern map (high level)\n\nThe README groups all **26 patterns** into six families:\n\n| Family | Examples | \n|---|---|\n| **A. Staging instead of stating** | Patterns 1–5 above | \n| **B. Rhythm by rule** | Forced triads, repeated openings, dash spam | \n| **C. Inflation and borrowed authority** | “Delve,” “testament,” vague associations, fake expert consensus | \n| **D. Formatting by rule** | Bold decoration, emoji headings | \n| **E. Chat and draft leftovers** | “Great question!” wrappers, knowledge-limit disclaimers | \n| **F. Wrong reader** | Re-explaining what the audience already knows | \n\nEach row in the README includes **before** and **after** micro-examples — useful when you tune your own **style guides** for agents.\n\n## How the skill runs (workflow)\n\nFrom **`SKILL.md`** and the README:\n\n1. **Mark tells** in the incoming text, strongest first.\n2. **Draft** a rewrite without treating the original outline as sacred.\n3. **Critique** the draft for remaining tells and factual drift.\n4. **Ship** final prose; for files, change**prose only** and leave code, data, frontmatter, and link targets alone.\n\nWhen you paste text interactively, the skill shows **first draft, critique, final** — transparency that helps humans learn patterns, not only outsource editing.\n\n**Voice matching:** paste **2–3 paragraphs of your own writing** and Humanizer will mimic rhythm and punctuation quirks, including dashes if your sample uses them.\n\n## Installation paths (copy from README)\n\n**Claude Code (plugin):**\n\n```\n/plugin marketplace add blader/humanizer\n/plugin install humanizer@humanizer\n```\n\nRequires **Claude Code 2.1.142+** per README; older clients fall back to Skills CLI.\n\n**Codex and multi-agent:**\n\n```\nnpx skills add blader/humanizer --global --agent codex\n```\n\n**Every agent the CLI knows:**\n\n```\nnpx skills add blader/humanizer --global --agent '*'\n```\n\n**Claude.ai / Desktop:** download ZIP and upload as a skill in settings — README path for non-CLI users.\n\nFor install hygiene, mirror checks from [ECC affaan-m harness skills](https://explainx.ai/blog/ecc-affaan-m-agent-harness-71-agents-311-skills-install-risks-github-trending-guide-2026): read **`SKILL.md`**, verify licenses, and know what network calls your agent may still make outside the skill.\n\n## What developers are saying on Hacker News\n\nAn older HN thread on Humanizer (**item 47109489**) is mostly normative, not benchmark-heavy. Commenter [davydm](https://news.ycombinator.com/item?id=47109503&ref=explainx) called it “another step in the war” they would not champion — a values objection to polishing AI prose. [wps](https://news.ycombinator.com/item?id=47109803&ref=explainx) lamented that automated polish drains meaning from writing. Those reactions are **not** hands-on throughput numbers; they warn that **humanizer culture** can collide with authenticity goals. Use the skill when **clarity** is the goal, not when **voice** is the product.\n\n## Limits you should plan for\n\nThe README is explicit:\n\n- **Detectors** may still flag output — do not use it for academic integrity evasion.\n- **Facts** must not be invented — it should**ask** when a rewrite needs missing data.\n- **Technical reference** stays neutral; personal essays keep quirks.\n\nHumanizer does not replace [citation checking](https://explainx.ai/blog/how-to-read-hugging-face-model-card-2026)-style discipline or legal review for regulated copy. Keep a human editor on the hook for numbers that appear only once in a draft — pattern stripping does not catch invented statistics.\n\n## Pattern deep dive — inflation and borrowed authority\n\nPatterns **12–18** in the README are where enterprise blogs fail reviews. The README calls out overused tokens such as **“delve,” “testament,”** and **“landscape,”** plus **inflated significance** lines like **“marking a pivotal moment.”** Enterprise risk teams increasingly grep for those words after compliance incidents.\n\nPattern **17 — borrowed authority** is especially toxic in B2B copy: **“Experts believe”** without naming an expert, or a **logo wall** of outlets with no linked study. Humanizer’s fix is blunt: **name a real source or delete the claim.** That aligns with explainx.ai’s guidance in [how to read model cards](https://explainx.ai/blog/how-to-read-hugging-face-model-card-2026) — citations must point somewhere inspectable.\n\nPattern **18** swaps **“serves as / features / boasts”** for plain **“is / has.”** It sounds trivial; it removes an entire class of **press-release passive voice** agents love.\n\n## Pattern deep dive — formatting and chat leftovers\n\nPatterns **19–25** catch **bold decoration**, **emoji headings**, and **chatbot residue** such as **“Great question!”** wrappers. Those lines often survive multi-agent pipelines when one model drafts and another formats.\n\nPattern **25 — writing about the document** shows up in agent PRs: **“This function was added to replace…”** Humanizer pushes the model to describe **behavior**, not **diff meta**.\n\n## Using Humanizer inside a daily agent loop\n\nA practical loop for doc teams:\n\n1. **Draft** with your coding agent under a normal system prompt.\n2. **Run**`/humanizer` on paragraphs that face customers or regulators — not on every code comment.\n3. **Diff** the prose-only changes; reject any fact edits you did not authorize.\n4. **Run** your existing**style linter** and**legal review** — Humanizer is not compliance.\n\nIf you orchestrate with [loop engineering patterns](https://explainx.ai/blog/loop-engineering-coding-agents-claude-code-guide-2026), add Humanizer as an **optional pass** after factual QA, not before technical QA. Order matters: humanizing hallucinations just produces **polite hallucinations**.\n\n## Comparison — what Humanizer does not replace\n\n| Tool or habit | Humanizer overlap | \n|---|---|\n| Manual copy edit | Overlaps — Humanizer automates pattern removal | \n| [AI verification tax](https://explainx.ai/blog/ai-verification-tax-rework-sage-workday-foxit-surveys-how-to-cut-it-2026) workflows | Complements — still need fact checks | \n| Plagiarism or rights scanners | No overlap | \n| [Claude Security scans](https://explainx.ai/blog/claude-mythos-5-security-scans-defender-advantage-fund-august-2026) | No overlap — security not voice | \n| Custom `SKILL.md` voice guides | Can combine — feed your samples for voice match | \n\n## Wikipedia maintenance context\n\nPen tracing a wavy line from stacked letters, echoing Wikipedia’s Signs of AI writing editorial guide\n\nWikiProject AI Cleanup does not maintain Humanizer, but both lean on the same [Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing?ref=explainx) page. Editors remove slop from articles; Humanizer removes slop from drafts **before** you paste into CMS tools.\n\nThat shared origin is a feature: when your comms team asks “why does this read like ChatGPT,” you can point to a **public pattern list** instead of taste arguments. It also means when Wikipedia updates the guide, skill authors may need to **merge upstream pattern changes** — watch the repo CHANGELOG.\n\n## Sponsorship and supply chain\n\nThe README discloses **cfo.ai** sponsorship. Sponsorship does not change the MIT license or the Wikipedia attribution. Still follow normal supply-chain hygiene from [ECC skills install guidance](https://explainx.ai/blog/ecc-affaan-m-agent-harness-71-agents-311-skills-install-risks-github-trending-guide-2026): read **`SKILL.md`**, check plugin marketplace signatures, and pin versions in team docs.\n\n## Ethics for teams\n\nIf you edit customer-facing text:\n\n- Disclose **AI assistance** where your policy requires it — rewriting does not remove machine origin.\n- Keep **human review** on numbers, security claims, and comparisons.\n- Do not use Humanizer to **launder** plagiarized or scraped content — pattern stripping is not rights clearance.\n\n## Related reading\n\n*Star counts and install commands match the blader/humanizer README as of November 8, 2026.*", "url": "https://wpnews.pro/news/blader-humanizer-wikipedia-pattern-skill-to-strip-ai-writing-tells", "canonical_source": "https://explainx.ai/blog/blader-humanizer-agent-skill-wikipedia-ai-writing-patterns-github-trending-guide-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-08 00:00:00+00:00", "updated_at": "2026-11-09 02:01:14.537702+00:00", "lang": "en", "topics": ["ai-tools", "ai-agents", "generative-ai", "developer-tools"], "entities": ["blader/humanizer", "GitHub", "Wikipedia", "Claude Code", "Codex", "WikiProject AI Cleanup", "cfo.ai", "explainx.ai"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/blader-humanizer-wikipedia-pattern-skill-to-strip-ai-writing-tells", "markdown": "https://wpnews.pro/news/blader-humanizer-wikipedia-pattern-skill-to-strip-ai-writing-tells.md", "text": "https://wpnews.pro/news/blader-humanizer-wikipedia-pattern-skill-to-strip-ai-writing-tells.txt", "jsonld": "https://wpnews.pro/news/blader-humanizer-wikipedia-pattern-skill-to-strip-ai-writing-tells.jsonld"}}