SkillFoxx

humanizer-stack skill

Using two Claude Code skills, the agent edits wording and structure, flagging stock phrases and repeated narrative patterns.

  • 382 ★ on the repository
  • Last commit
Skill

Extended access

Why this level

  • Local Python scanners read drafts and the agent may edit their files; review changes before publication.
All reasons and checks

nulightjens/humanizer-stackAuthor:NulightJens

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/humanizer-stack

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Local Python scanners read drafts and the agent may edit their files; review changes before publication.

This is third-party code. Review the repository files before installing.

What it does

With humanizer-stack, the agent edits in two passes. It first checks wording and phrasing, then reviews structure: explicit morals, overly tidy progression and resemblance to earlier pieces. The agent selects a limited number of structural changes for each text. Two Python scanners find pattern-matchable signals. They help collect evidence, while the editor decides whether to reorganize paragraphs.

Who it is for. Newsletter and blog editors whose drafts retain repetitive structures after sentence-level editing.

Good fit when

  • After wording cleanup, an article still follows the same shape as the previous three pieces.
  • An editor wants separate checks for phrasing and the progression of a long draft.
  • A team needs a machine-readable report of detected writing patterns.
  • A lesson needs repetitive moral-of-the-story endings removed.

Not a fit when

  • The task requires proving authorship from scanner output.
  • An editor expects guaranteed results against an external AI detector.
  • You want every structural intervention applied automatically to each short message.

Example request

Edit this newsletter’s phrasing first, then compare its structure with the previous two editions. Propose only changes you can justify.

Limitations

The structural guidance draws on research about roughly 5,000-word fictional stories; applying it to short business copy requires editorial judgment. Scanners only catch formal patterns and do not verify facts, rhythm or the writer’s viewpoint. The installer creates skill symlinks by default, so moving the repository breaks those connections. Original work uses MIT, but incorporated material has separate terms, including CC BY-SA 4.0.

How to disable. Remove humanizer and structural-humanizer from Claude Code’s skills directory. Check whether installation used symlinks or independent copies before deleting the repository.

FAQ

Which skill runs first?

Run humanizer before structural-humanizer. After editing wording, the agent examines the structure.

Can I get a report for automation?

Yes. `python3 scripts/copy_scan.py --json draft.md` returns machine-readable output. Review findings before automatically changing prose.

Should all six structural checks become edits?

No. The checks identify issues; choose one or two justified interventions for the particular piece to avoid giving different texts the same new shape.

How can I install independent skill copies?

Pass `--copy` to the installer. Repository updates will then leave the installed copies unchanged.

Can I keep an intentional pattern flagged by the scanner?

Yes. A `copy-ignore` marker suppresses that line in the copy scanner. Use it after an editorial decision.

Discussion

No questions or reviews yet.

Skills

A skill that finds AI writing tells, rewrites the text in a chosen voice, and optionally scores it from 0 to 100

SkillMinimal access311
Russian stack

An editor for AI-written Russian text: removes bureaucratese, cliches and 38 AI patterns while keeping facts intact

SkillMinimal access210
Russian stack

A Russian text editing skill: removes bureaucratese, calques and AI mannerisms without touching facts, with a scanner to measure the edit

SkillExtended access562

A skill that edits German AI drafts: it removes stock phrases, keeps facts and the author's voice and checks 72 language patterns.

SkillMinimal access205

Two skills for English prose: humanize rewrites a model draft in a natural voice, and ai-check flags machine-writing tells

SkillMinimal access622
Editors’ pick

A skill that strips AI writing tells from text using Wikipedia's list, without adding invented facts

SkillMinimal access68.4K

README in short

Separate references cover public-copy patterns and genre calibration. A pipeline guide assigns responsibilities to the two passes.

README badge

Original SKILL.md

---
name: structural-humanizer
description: >-
  Remove the discourse-level (structural) signs of AI writing that survive surface
  editing: stated lessons and moral-of-the-story closers, tidy single-track arcs,
  embodied-emotion performance ("chest tightened"), vague allusions instead of named
  references, unbroken linear structure, and shape convergence across pieces. Grounded
  in the StoryScope study (Russell et al. 2026): narrative structure alone detects AI
  text at 93.2% F1, and professional stylistic rewriting moved detection only 1.6
  points. Use as the SECOND pass after the humanizer skill (which handles words and
  phrasing) whenever writing or revising LinkedIn posts, course lessons, blog posts,
  essays, newsletters, or emails that must read as human. Triggers: "humanize",
  "de-slop", "AI tells", "make this sound human", "structural pass", "deep humanize".
---

# structural-humanizer

Read this first. The `humanizer` skill fixes words: "delve", em dashes, rule of three,
negative parallelism. This skill fixes what survives that pass: the structure. The two
are different jobs, run in sequence. Surface pass first, structural pass second.

**Why this layer matters more.** StoryScope (Russell et al. 2026, arXiv:2604.03136)
classified 61,608 stories from humans and 5 LLMs using only discourse-level features,
with all style features withheld: 93.2% detection accuracy. Then they ran AI text
through LAMP, a professional span-level rewriting framework that removes cliche,
purple prose, and redundant exposition (functionally, a surface humanizer). Detection
dropped 1.6 points. Meanwhile the surface layer is decaying on its own: GPT 5.4
already slashed em-dash usage, and fine-tuning drops stylistic detection from 97% to
3%. The durable fingerprint is structural, and fixing it requires structural rewrites,
not word swaps. Full findings with numbers: [references/storyscope-findings.md](references/storyscope-findings.md).

## The trap (same trap as unslop-ui)

Do not replace one default with another. If every piece now opens mid-scene, names
three feelings, and ends unresolved, that is a new detectable cluster. The study's
deepest finding is convergence: all five AI models occupy one tight region of
structural space while humans are dispersed and rare. Rarity IS the human signal.

So: **pick 1-2 structural interventions per piece, vary them across pieces, and be
able to say why this piece got this shape.** Never apply the whole menu at once.

## The six audits

Report a problem