tizzy916/humanities-writing-companion

What it is

Humanities Writing Companion is an Agent Skill – a plug‑in written in the open SKILL.md format that can be run by Claude Code, the Claude Agent SDK, or any other AI‑agent platform that understands the format. It acts as a virtual “thinking partner” for scholars in the humanities, guiding them through every stage of a long‑form argumentative paper – from sharpening the research question to generating a journal‑ready AI‑use disclosure.

Who it’s for

The skill targets researchers whose primary output is prose‑driven argumentation: historians, philosophers, literary critics, art historians, religious studies scholars, classicists, cultural‑studies scholars, and related interdisciplinary fields. It is not a generic grammar‑checker or citation manager; it is built around the specific scholarly values of voice preservation, argument‑through‑prose, and adversarial peer review.

Core capabilities (the 12 modes)

Stage	Mode	What it does
Pre‑writing	H – research‑question sharpening	Socratic dialogue that turns a vague interest into a sharp, write‑able question.
	I – literature‑mapping	Organises existing readings into camps and debates (no automated literature search).
	J – plan‑only outlining	Generates discipline‑aware paper structures without writing prose.
Drafting	C – conception → new content	Co‑writes new sections following the author’s voice profile.
	A – paragraph‑level dialogue	Interactive edit of a single paragraph, preserving stylistic quirks.
Review	B – chapter‑level review	Four‑layer critique (foundation, structure, paragraph, sentence) with priority‑aware feedback.
	D – devil’s‑advocate	Simulates three reviewer archetypes (theoretical, historical, methodological) plus a friendly reader, with adjustable intensity.
Revision	E – writing‑bottleneck assistance	Suggests unblock strategies when the author is stuck.
	F – draft revision	Side‑by‑side diff, anti‑AI‑cliché checks, and a “revision‑coach” sub‑mode.
Pre‑submission	G – blind reading	Checks for hidden assumptions and readability for an anonymous audience.
	K – AI‑use disclosure	Audits all AI contributions and produces a journal‑ready disclosure statement.
Post‑review	L – revision workflow	Organises reviewer comments into a structured dossier and tracks integration status.

Supporting tools

The repository ships a scripts/ folder (zsh + Python 3, no external packages) that provides:

• ai‑trace‑scan.sh – flags AI‑generated clichés and repetitive transitions.
• pending‑checks.sh – aggregates all [VERIFY] / [待核对] markers for a final clean‑up.
• citation‑consistency.py – validates citation‑style consistency (Chicago, MLA, APA, GB/T 7714).
• citation‑format‑convert.py – converts a BibTeX file between the supported styles.
• citation‑verify.py – looks up Crossref/OpenAlex metadata for bibliography entries (no claim verification).

How to get it

# Install as a Claude Code skill
git clone https://github.com/tizzy916/humanities-writing-companion.git \
  ~/.claude/skills/humanities-writing-companion
chmod +x ~/.claude/skills/humanities-writing-companion/scripts/*.sh

Or clone into a project‑local .claude/skills/ folder for use with a single vault.

Why it matters

Most AI‑writing assistants treat humanities papers like any other text: they polish grammar, suggest citations, or rewrite for “standard academic English.” This skill recognises three humanities‑specific priorities that generic tools ignore:

1. Voice preservation – the author’s rhetorical stance is part of the argument.
2. Argument‑prose inseparability – a weak sentence is an invalid argument.
3. Adversarial peer review – the skill simulates the harsh, method‑specific critiques scholars face before submission. By embedding discipline‑aware critique strategies, version‑style revision logs, and an explicit AI‑trace checklist, the Companion helps scholars produce defensible, transparent, and personally authentic manuscripts.

---

Citation If you use the skill in a published work, cite the Zenodo DOI: 10.5281/zenodo.20280772.