humanities-writing-companion
End-to-end humanities writing assistant — an Agent Skill (open SKILL.md format). 11 modes from Socratic research-question sharpening through AI-use disclosure. Bilingual (EN/中文), discipline-aware (literature/history/philosophy/art/religion/linguistics + cross-disciplinary). Four-layer critique, calibratable devils advocate, voice preservation.
原始仓库简介 · 中文摘要待补充 · academic-writing / bilingual / claude-code / claude-skill / claude-skills / devil-advocate
README / 项目原文
Humanities Writing Companion · 人文学科写作伙伴
An agent skill for humanities scholars whose primary deliverable is a long-form argumentative text — history, philosophy, literature, cultural studies, art history, religious studies, classics. Written in the open Agent Skills (SKILL.md) format — works with Claude Code, the Claude Agent SDK, and any agent that supports the format.
-orange){kind=icon}
📖 Wiki · 中文版 README · Skill source · English · Skill source · 中文
Table of contents
- Positioning
- What this skill takes seriously
- A typical interaction
- Core features
- Supported humanities disciplines
- Showcase: Before / After
- Install
- Quick start · three typical scenarios
- Comparison with adjacent tools
- Project structure
- Design philosophy
- Citation
- Contributing
- About the Author
- License
- Acknowledgments
Positioning
End-to-end writing assistant for humanities scholars — covering the full lifecycle of a humanities paper from research question to submission disclosure:
research question → literature map → planning → drafting → revision →
adversarial review → AI-trace cleanup → blind-reading check → AI-use disclosure
Built for fields where prose IS the argument — history, philosophy, literature, cultural studies, art history, religious studies, classics, intellectual history, science studies, and adjacent humanities-aligned fields.
Not a polishing tool. Not a citation manager. Not a research pipeline. A thinking partner that stays with you across the whole arc.
12 modes covering the writing lifecycle
| Stage | Modes |
|---|---|
| Pre-writing | Mode H · Research-question sharpening · Mode I · Literature mapping · Mode J · Plan-only outlining |
| Drafting | Mode C · Conception → new content · Mode A · Paragraph dialogue |
| Review | Mode B · Chapter review (4-layer critique) · Mode D · Devil's advocate (calibratable 1–5 + methodology-focus) |
| Revision | Mode E · Writing-bottleneck assistance · Mode F · Draft revision (with revision-coach sub-mode) |
| Pre-submission | Mode G · Blind reading · Mode K · AI-use disclosure |
| Post-review | Mode L · Revision workflow (defense/external-review comment integration, revision-dossier system) |
Plus a citation toolchain (scripts/): consistency checking, format conversion (Chicago / MLA / APA / GB7714), and metadata candidate lookup through Crossref/OpenAlex (not source-text or claim verification). In agent-capable environments (e.g., Claude Code), Mode B/D reviews can fan out to parallel reviewer agents, and pending claims can be checked against sources with provenance, support, and limitations recorded separately.
What this skill takes seriously
Three things generic AI writing tools do not handle for humanities:
1. Voice preservation is a core scholarly value, not stylistic decoration. In humanities, the author's voice carries epistemic weight: it signals which intellectual tradition the author writes from, which interlocutors they take seriously, which moves are theirs and which are borrowed. A paper polished into "standard academic English" loses this signal. This skill helps you write more like yourself, not less.
2. Argument and prose are inseparable. In empirical research, you can have a perfect experiment ruined by bad writing. In humanities, the writing IS the argument — a slack sentence, a vague concept, an unwarranted transition is an argumentative failure. This skill works at the level of argument-through-prose.
3. The reviewer is real and adversarial. Humanities reviewers are not gentle. A theoretical concept will be tested for sharpness; a historical claim will be tested for evidence; a philosophical argument will be tested for the strongest counter. This skill simulates that adversary internally so the paper meets it before submission.
A typical interaction
You: 帮我看看这段。我在论福柯的全景敞视主义如何延伸到数字平台。
[skill reads the paragraph + your style profile]
Companion:
I notice three things, in order of priority.
🔴 Foundation: you're using "全景敞视" as a metaphor for platform
surveillance, but Foucault's original concept turns on architecture
producing a specific subject-position. Platforms produce a different
subject (data subject, not body in a cell). You may want to either
(a) name the analogy as analogy and discuss the difference, or
(b) borrow Zuboff's "instrumentarian power" which fits the data case
more cleanly. Which direction do you want to go?
🟡 Argument: paragraph 2 jumps from "platforms collect data" to "this
constitutes surveillance" without the middle step about visibility-
as-power. A friendly reader will not follow.
🟢 Voice: your style profile has "first-person with em-dash insertion."
Paragraph 3 starts with "本文认为…" which reads like cliché. Want to
restore the first-person?
You: 走方向 (b). 改吧。
[skill produces revision; you keep the version that sounds like you]
This is what "thinking partner, not polishing tool" means in practice.
Core features
Four-layer critique (not single-axis polishing)
Layer 1 · Foundation "Does this paper hold up scholarly?"
Layer 2 · Structure "How is the argument unfolding?"
Layer 3 · Paragraph "What is this paragraph doing?"
Layer 4 · Sentence "Is this sentence right? Well-said?"
Scope-aware priority: start with consequential argument issues in a full review; complete a requested local edit when it does not depend on resolving those issues.
Devil's advocate with anti-sycophancy
Simulates three reviewers + one well-intentioned reader:
- Reviewer A · Theoretically demanding
- Reviewer B · Historically empirical
- Reviewer C · Methodologically skeptical
- Reader D · Well-intentioned but confused (distinctive design: places where a friendly reader gets confused are weak points in the argument)
Evidence-based correction: do not concede merely under pressure, but retract a challenge when even one decisive source passage or valid argument defeats it. Reviewer intensity never permits ignoring a correction.
Deep voice learning and preservation ("My hand writes my voice")
Not just sentence preferences. Also:
- Argumentative rhythm (linear / spiral / variable-tempo)
- Scholarly posture (critical inheritance / dialogical advance)
- Rhetorical function of citation (authority anchor / critical target / dialogue interface / narrative / conceptual tool)
- AI-trace checklist (8 categories of unexamined patterns ≠ AI cliché ≠ scholarly cliché)
Discipline-specific dimensions
Different chapter types get different critique strategies:
- Historical narrative · anachronism, counterfactual stress, source handling
- Philosophical argument · conceptual derivation, cross-theoretical transplantation, strongest objection
- Literary criticism · close reading vs. interpretation, genre awareness, form and meaning
- Cultural studies · power-knowledge framing, positionality, generalization range
- Art history · description vs. interpretation, provenance, reception
- Religious studies / classics · source-language rigor, tradition awareness, insider-outsider position
- Case analysis (cross-disciplinary)
ADHD-friendly interaction
- Feedback batched 3-5 per round
- Quick wins first
- Topic-jump support (a jump may be an insight signal)
- Pomodoro-friendly task units
- Reorientation summaries every 4-5 turns
Reflexive writing support (distinctive module)
If your research itself involves human-AI collaboration (autoethnography of AI-assisted writing), the skill provides six categories of "reflexive moments":
🔄 Direction change · 🚫 Refusal · 🎭 Voice conflict · 🔧 Tool dependency · 💡 Unexpected insight · 🤖 AI-trace awareness
Scholarly basis (citable in your paper):
- Christou (2026). Reconfiguring Reflexivity in the Era of AI. Qualitative Inquiry.
- Wiles (2025). Recursive Cognition in Practice. International Journal of Qualitative Methods.
- Panke (2025). How Can (A)I Research This? An Autoethnographic Exploration of Generative AI.
Engineering infrastructure
Borrowed from software engineering, in service of humanities writing:
- Version management: minor version = commit / major = release /
_drafts/= feature branch - Revision log: every change recorded with diff + reason (like git commit message)
- Feedback reports: Blocker / Major / Minor / Question tiers (from code review)
- Systematic verification checklists: argument completeness / concept consistency / citation completeness / style consistency
[VERIFY]/[待核对]hard marker: anti-citation-hallucination — cannot enter submission version
Twelve work modes (not one)
- Mode A — paragraph-level dialogue
- Mode B — chapter-level review
- Mode C — conception → new content writing (with collaborative drafting protocol)
- Mode D — devil's advocate (calibratable 1–5, methodology-focus sub-mode)
- Mode E — writing bottleneck assistance (5 unblocking strategies)
- Mode F — draft revision (two-version comparison, anti-AI-cliché, revision-coach sub-mode)
- Mode G — blind reading (mechanical promise-delivery check)
- Mode H — research-question sharpening (Socratic dialogue: from vague interest to a sharp, write-able question, with "so what" test and interlocutor identification)
- Mode I — literature mapping (organizes what you have already read into camps and debates — never searches literature for you)
- Mode J — plan-only outlining (discipline-aware standard arcs; plans the paper without writing prose)
- Mode K — AI-use disclosure (audits actual AI involvement, 4-tier categorization, journal-ready disclosure statement)
- Mode L — revision workflow (defense/external-review comment integration: one revision dossier per comment, status-authoritative master table)
Engineering helper scripts
scripts/ provides five tools requiring zsh and Python 3, with no third-party Python packages:
| Script | Purpose |
|---|---|
ai-trace-scan.sh |
Scan clichés and transition pile-ups |
pending-checks.sh |
Aggregate all [VERIFY] / [待核对] / ❓ to discuss / [AI DRAFT] markers |
citation-consistency.py |
Citation-format consistency check (brackets / commas / connectors / EN/CN names / page numbers) |
citation-format-convert.py |
Convert a BibTeX bibliography between Chicago (Author-Date) / MLA 9 / APA 7 / GB/T 7714 |
citation-verify.py |
Look up Crossref/OpenAlex metadata candidates (FOUND / FUZZY_MATCH / NOT_FOUND / ERROR); no source-text or claim verification |
Supported humanities disciplines
This skill organizes humanities scholarship in a three-layer architecture, so the discipline-routing system can match where the author actually works — not just to one of a flat list of seven slots. Authors declare their discipline at onboarding (or the skill infers from the draft); routing then loads the matching layer.
L1 · Six humanities main disciplines
These are the canonical L1 humanities disciplines. Each carries a core set of methodological concerns generic AI writing tools miss.
| L1 discipline | Object of study | Core methodological concerns |
|---|---|---|
| Literature · 文学 | Texts (poetry, fiction, drama, essay) | Close reading vs. interpretation · Genre awareness · Form-meaning fit · Intertextuality |
| History · 史学 | Past events, persons, societies | Anachronism · Counterfactual stress · Source handling (primary vs. secondary) · Causal-chain transparency · Historiographical positioning |
| Philosophy · 哲学 | Concepts, arguments, normative claims | Conceptual derivation · Argument form (formal vs. material) · Cross-theoretical transport cost · Steel-manning the strongest objection · Modal scope |
| Linguistics · 语言学 | Language structure and use | Data source (corpus vs. intuition vs. elicitation) · Form vs. function · Description vs. prescription · Cross-linguistic claim scope |
| Art studies · 艺术学 | Art works (painting, sculpture, music, film, architecture) | Description vs. interpretation (keep separate) · Provenance and materiality · Reception history · Medium-specific form analysis |
| Religious studies · 宗教学 | Religious traditions, texts, practices | Source-language rigor (original vs. translation) · Tradition position · Insider/outsider (emic vs. etic) · Comparative method |
L2 · Common subfields (not exhaustive)
Subfields inherit all the methodological concerns of their parent L1, plus any specific constraints the author declares at onboarding. Examples — many more possible:
| Parent L1 | Example subfields |
|---|---|
| Literature | Classical Chinese literature · Modern Chinese literature · Comparative literature · Literary theory · Literary criticism · Foreign-language literatures |
| History | Chinese history · World history · Economic history · Social history · Cultural history · Urban history · Periodized fields (Tang history, early modern Europe, etc.) |
| Philosophy | Chinese philosophy · Western philosophy (analytic vs. continental) · Ethics · Aesthetics · Political philosophy · Philosophy of science · Phenomenology |
| Linguistics | Historical linguistics · Sociolinguistics · Pragmatics · Typology · Discourse analysis |
| Art studies | Art history · Musicology · Film studies · Theatre studies · Architectural history |
| Religious studies | Christian studies · Buddhist studies · Daoist studies · Comparative religion |
If your subfield isn't listed, declare it at onboarding — it inherits from its parent L1 automatically.
L3 · Cross-disciplinary fields (explicit multi-inheritance)
These are humanities fields that explicitly draw from multiple L1s. The skill loads the methodological concerns of all parent L1s plus the L3-specific overlay.
| L3 field | Inherits from | L3-specific overlay |
|---|---|---|
| Cultural studies · 文化研究 | Literature + History + Sociology | Power/knowledge framing · Positionality · Generalization range |
| Classics · 古典学 | Literature + History + Philosophy + Religious studies + Archaeology | Textual criticism (manuscript tradition) · Philological rigor · Reception history |
| Intellectual history · 思想史 | History + Philosophy | Begriffsgeschichte vs. Cambridge School · Context vs. text · Avoiding presentism |
| History of science · 科学史 | History + Science + Philosophy | Internal vs. external history · Whig-history vigilance · Technical accuracy · Case-study calibration |
| Media studies · 媒介研究 | Literature + Cultural studies + Philosophy of technology | Medium-morphology · Reception studies · Tech-social co-construction |
| Digital humanities · 数字人文 | Any L1 + Computation | Data reproducibility · Tool transparency · Algorithmic bias · Methodological disclosure of computational choices |
| Gender studies · 性别研究 | Literature + History + Cultural studies | Gender ontology · Historicizing gender · Intersectionality |
| Postcolonial studies · 后殖民研究 | Literature + History + Cultural studies | Positionality · Translation politics · Resisting Eurocentrism |
| Environmental humanities · 环境人文 | Literature + History + Science | Anthropocene framing · Multi-species perspective · Scale problems (local vs. planetary) |
Humanities-adjacent fields (welcome with scope note)
Some fields are formally classified as social science but include strong humanities-style sub-traditions (where prose IS the argument). This skill welcomes such work:
| Field | What we serve | What we don't serve |
|---|---|---|
| Communication studies · 传播学 (humanities-style) | Media ecology school (Innis / McLuhan / Postman / Carey); critical communication; humanities-aligned cultural-media research | Empirical/experimental communication research, content analysis as quantitative method |
| Educational research · 教育学 (humanities-style) | History of education · Philosophy of education · Curriculum theory in humanities mode · Critical pedagogy | Quantitative educational psychology · Empirical learning-outcomes research |
Inheritance: communication studies (humanities-style) ← Media studies + Philosophy + Cultural studies. Educational research (humanities-style) ← History + Philosophy + Cultural studies.
Fallback protocol — if you still don't see your field
Tell the skill at onboarding:
- Object of study: text / past event / concept / phenomenon / artifact / practice / language structure / …
- Primary method: close reading / archival research / argument analysis / ethnography / comparison / philological work / …
The skill will infer your closest L1 + any relevant L3 overlays, write it to _writing-config/discipline.md, and proceed. You can refine the declaration anytime.
The full methodology rubric for each L1 / L3 entry lives in references/disciplines.md — this README table is the surface map; the skill carries the full rubric.
Showcase: Before / After
An illustrative exercise, not a validated quotation from a named scholar. The supplied material is deliberately limited:
Draft. All apprentices learn primarily from written rules. The two workshop interviews demonstrate that formal rules always take priority over experience. Reading note supplied by the author. In these two interviews, apprentices describe consulting written rules for routine tasks and asking experienced workers about unusual cases.
The skill should identify the mismatch before strengthening the prose:
Finding · claim–evidence mismatch. The interviews describe two forms of learning; they do not establish that rules always take priority. Two workshop interviews also do not justify “all apprentices.” This judgment uses the supplied note; the interview records have not been checked.
If the author requests a revision, a bounded proposal is:
After. In the two interviews summarized here, apprentices describe using written rules for routine tasks and turning to experienced workers in unusual situations. These accounts suggest that the roles of rules and experience vary with the task. [VERIFY: check the original interview records and add their actual locators.]
The revision narrows the claim to the material supplied and makes the remaining source check visible. It does not invent a quotation, page number, or external scholarly authority. The author still decides whether this interpretation fits the full evidence.
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 as a project-level skill (vault / project only):
git clone https://github.com/tizzy916/humanities-writing-companion.git \
./.claude/skills/humanities-writing-companion
Claude Code loading
Claude Code auto-scans ~/.claude/skills/ and ./.claude/skills/ on startup. After install, say "I'm working on a humanities paper" or any of the trigger phrases below.
Without git (ZIP download)
No git required: on the GitHub repository page, click Code → Download ZIP, unzip, and move the unzipped folder to ~/.claude/skills/humanities-writing-companion (or ./.claude/skills/humanities-writing-companion for project-level install). Then make the shell scripts executable:
chmod +x ~/.claude/skills/humanities-writing-companion/scripts/*.sh
Claude desktop app / claude.ai
The Claude desktop app and claude.ai also support custom skills: package the skill folder (the directory containing SKILL.md) as a .zip and upload it in your Claude settings, under the capabilities/skills section (the exact menu wording may vary as the product evolves — look for "Skills"). Note that the scripts/ toolchain requires a shell-capable environment (Claude Code / agent mode); in chat-only environments the skill's dialogue modes work, but scripts do not run.
Claude Agent SDK
SKILL.md can be loaded into your system prompt directly. The skill is plain text — no runtime dependencies.
Other agents (open SKILL.md format)
This skill is written in the open Agent Skills format: a folder containing SKILL.md plus plain-text references/ and scripts/. Any agent that supports the format — or that can simply read SKILL.md (and, when routed there, references/*.md) into context — can use it: clone this repo into wherever your agent discovers skills. The scripts/ toolchain requires zsh and Python 3. Citation lookup additionally needs network access to Crossref and OpenAlex; the other checks run locally.
Verify the installation
After installing, start a new conversation and either:
- Ask Claude: "What skills do you currently have loaded?" —
humanities-writing-companionshould appear in the list; or - Say a trigger phrase directly, e.g. "review my section" or "帮我看看这段" — the skill should activate and respond in its four-layer-critique voice rather than as a generic polisher.
If neither works, check that the folder sits directly under ~/.claude/skills/ (i.e., ~/.claude/skills/humanities-writing-companion/SKILL.md exists) and restart Claude Code.
Trigger phrases
English: "paper," "essay," "chapter," "dissertation," "argument," "thesis," "revise," "voice," "review my section," "stuck on writing," "devil's advocate," "reviewer attack"
Chinese: 论文 · 写作 · 润色 · 改论文 · 帮我看看这一章 · 我手写我口 · 这个论证有没有问题 · 我写不下去了 · 审稿人会怎么攻击
When an academic draft is already in context, casual requests such as "take a look at this paragraph" · 帮我看看这段话 also apply.
Quick start · three typical scenarios
Scenario 1: new paper, first use
Say to Claude: "I want to write a paper on X."
The skill starts from your idea and any materials already supplied, helping you form a research question or requested draft. It asks only for consequential missing information. Profiles and folders are created as needed when you choose ongoing file-based work.
Scenario 2: revising an existing chapter
"Help me read this chapter" → Mode B (chapter review) → 4-tier feedback report
"Help me revise this paragraph" → Mode A (paragraph dialogue) → diagnose + suggest + reason
"I'm stuck" → Mode E (bottleneck) → one suitable next move
Scenario 3: fighting AI cliché (two-version comparison)
If your paper has been through AI polishing but you want to restore the original voice:
Mode F · draft revision → compare AI-polished vs. original → keep improvements + restore voice
Comparison with adjacent tools
| Tool | Their focus | Where this skill differs |
|---|---|---|
| Jenni AI | Real-time autocompletion + literature search | We do thought-dialogue, not autocompletion. Real-time prediction skips the cognitive work that humanities argument needs. |
| Paperpal | Academic language polishing (STEM/biomed-leaning) | We're a writing architecture (12 modes, 4-layer critique, discipline routing), not a point polishing tool. |
| Yomu AI | Sourcely literature engine + paragraph feedback | We assume the author manages literature (Zotero/Drive). Mode I helps organize what you've already read — never replaces the reading. |
| Thesify | Paper Digest + Purpose-Check | Mode G is inspired by Purpose-Check. We use it within a broader four-layer critique workflow plus reviewer calibration. |
| HyperWrite Devil's Advocate | Point-tool counter-argument generation | Mode D is a full devil's-advocate mode with 1–5 calibration, methodology-focus sub-mode, and evidence-based concessions (anti-sycophancy). |
| Grammarly / DeepL Write | Grammar / translation polishing | We never rewrite for "clarity" at the cost of voice. "My hand writes my voice" is a core principle, not optional. |
| Generic ChatGPT / Claude (no skill) | General-purpose chat | We carry persistent style profile, reader profile, revision log, four-layer critique, discipline routing, AI-trace checklist, and citation toolchain across sessions. |
Project structure
humanities-writing-companion/
├── SKILL.md ← Core skill file (EN: scope, evidence rules, task router, compact critique)
├── SKILL.zh.md ← Chinese mirror (中文版)
├── references/ ← On-demand manuals (each with a `.zh.md` Chinese mirror)
│ ├── critique-review.md ← Detailed four-layer questions and review coverage
│ ├── disciplines.md ← Full discipline dimension tables (L1/L2/L3/adjacent + fallback)
│ ├── modes-prewriting.md ← Mode H / I / J full protocols
│ ├── mode-c-drafting.md ← Mode C four-stage drafting flow
│ ├── mode-d-adversarial.md ← Mode D devil's-advocate full protocol
│ ├── mode-e-bottleneck.md ← Mode E bottleneck strategies
│ ├── mode-f-revision.md ← Mode F revision workflow
│ ├── modes-submission.md ← Mode G / K full protocols
│ ├── deep-style.md ← Deep style understanding & preservation
│ ├── multilingual-writing.md ← Mixed-language writing norms
│ ├── style-profile-template.md ← Style profile ("constitution of voice") template
│ ├── ai-trace-checklist.md ← AI-trace scan checklist
│ ├── project-management.md ← Project folder + version management
│ ├── revision-workflow.md ← Mode L revision-dossier workflow manual
│ └── target-reader-profile-template.md ← Target reader profile template
├── scripts/ ← Helper scripts (zsh + Python 3)
│ ├── README.md ← Script usage
│ ├── ai-trace-scan.sh ← AI cliché scan (zsh)
│ ├── pending-checks.sh ← Pending marker aggregation (zsh)
│ ├── citation-consistency.py ← Citation format consistency (Python 3)
│ ├── citation-format-convert.py ← Chicago/MLA/APA/GB7714 converter (v4.0+)
│ └── citation-verify.py ← Crossref/OpenAlex metadata candidate lookup
├── README.md ← This file
├── README.zh.md ← 中文 README
├── CHANGELOG.md ← Version history
├── LICENSE ← CC BY-NC 4.0
└── CITATION.cff ← Academic citation metadata
Bilingual status: the project is fully bilingual. SKILL.md, README, CONTRIBUTING, the references/ manuals, and scripts/README each exist as an English file plus a .zh.md Chinese mirror; script comments are bilingual as well. Both languages of trigger work either way (the description field in SKILL.md handles both).
Design philosophy
"My hand writes my voice"
Academic rigor and personal expression are not opposites. "Standard academic prose" usually means the death of individuality. The skill helps the author speak in their own voice rather than pressing their words into a prefabricated mold.
Thought first, format second
Revision priority:
- Force of the argument
- Precision of concepts
- Effectiveness of structure
- Quality of expression
- Format compliance
Always top-down. Do not fuss with commas in a paragraph whose underlying argument is broken.
Engineering rigor, humanistic expression
Borrows software engineering best practices (version management, unit tests, code review) in service of humanities writing. Engineering rigor does NOT mean turning the paper into code — it means every revision is traceable, argument quality is verifiable, the writing process is resumable, and problems are processed in layers.
Citation
If your research uses this skill, please cite it in the methodology section.
BibTeX:
@software{shen_humanities_writing_companion_2026,
author = {Shen, Cong},
title = {Humanities Writing Companion: An Agent Skill for Voice-Preserving Humanities Academic Writing},
year = {2026},
publisher = {Zenodo},
version = {5.1.0},
doi = {10.5281/zenodo.20280772},
url = {https://doi.org/10.5281/zenodo.20280772}
}
Plain-text attribution (for skill metadata, footers, etc.):
Based on Humanities Writing Companion by Shen Cong
https://github.com/tizzy916/humanities-writing-companion
See CITATION.cff for full machine-readable metadata (GitHub's "Cite this repository" button will use it automatically).
Citing companion tools
If you also use academic-research-skills in the same project, please cite both. ARS attribution format (per CC BY-NC 4.0):
Based on Academic Research Skills by Cheng-I Wu
https://github.com/Imbad0202/academic-research-skills
Contributing
Issues and PRs welcome:
- New work-mode proposals
- Extensions to the AI-trace checklist
- Discipline-specific examples (medieval studies, art conservation, ethnomusicology, etc.)
- Additional citation-format support (APA / Chicago / MLA / GB/T 7714 / journal-specific)
- Translation-quality improvements to the bilingual mirrors (
references/,scripts/)
See CONTRIBUTING.md.
About the Author
Shen Cong — BFA, Experimental Art, Central Academy of Fine Arts (CAFA); MA, History of Science, Tsinghua University (advisor: Hu Yilin); Founder & CEO of Tianyu Vision, a sci-art studio working on scientific visualization, science communication, and sci-art convergence.
This skill came out of writing the author's own MA thesis, Technical Liberalism. He noticed that almost every AI writing tool on the market pulled toward polishing and averaging — whereas humanities scholarship needs the opposite: protecting the author's scholarly voice, stress-testing argumentative rigor, and surviving adversarial peer review. So he built this skill — not to write for him, but to read for him, delivering at each of four layers (basic rigor / argument development / paragraph function / sentence-level expression) the kind of critique a real humanities scholar would actually give.
📮 GitHub @tizzy916 · shencong916@gmail.com · Corrections, collaboration, and conversation welcome.
License
CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International) — free for non-commercial use, modification, and distribution. Requires attribution.
⚠️ License change (v3.0.0, 2026-05-19): This project relicensed from MIT to CC BY-NC 4.0. Versions ≤ v2.1.0 remain under MIT and retain their original commercial-use rights for those specific versions. From v3.0.0 onwards, commercial use is prohibited without a separate license.
Commercial Use
This skill is licensed under CC BY-NC 4.0 — non-commercial use only (academic research, teaching, personal projects, open-source derivatives, internal research workflows).
For commercial licensing inquiries — embedding in a paid product, paid consulting or editing services using this skill, commercial SaaS integration, agency use on behalf of paying clients — contact the author for a commercial license:
📮 shencong916@gmail.com (Shen Cong · Tianyu Vision)
The author retains the right to grant commercial licenses on a case-by-case basis. Citing this skill in academic publications is always permitted regardless of license tier.
Acknowledgments
Methodological inspiration and scholarly basis:
- Christou, P. A. (2026). Reconfiguring Reflexivity in the Era of AI. Qualitative Inquiry.
- Wiles, F. (2025). Recursive Cognition in Practice. International Journal of Qualitative Methods.
- Panke, S. (2025). How Can (A)I Research This?.
- Foucault, M. (1984). What is Enlightenment? — "diagnostic of the present" as methodological tradition
- Stiegler, B. (2013). What Makes Life Worth Living: On Pharmacology — "critical pharmacology"
Some design patterns inspired by:
- Imbad0202/academic-research-skills — the upstream pipeline this skill complements; the Concession Threshold pattern in their reviewer module inspired Mode D's "minimum standard before conceding"
- Voice DNA + Audience Profile pattern — inspired the style profile + reader profile pairing
- Thesify Purpose-Check — inspired Mode G blind reading
人文学科写作伙伴 · Humanities Writing Companion
一个为人文学者设计的 agent skill，服务于以长篇论证性文本为主要交付物的写作——历史、哲学、文学、文化研究、艺术史、宗教学、古典学。采用开放的 Agent Skills（SKILL.md）格式——Claude Code、Claude Agent SDK 以及任何支持该格式的 agent 都可以使用。
📖 Wiki 教程 · English README · Skill 源文件 · 英文 · Skill 源文件 · 中文
目录
- 定位
- 本 skill 严肃对待的三件事
- 一次典型的交互
- 核心特性
- 支持的人文学科
- Showcase · 真实 Before / After 案例
- 安装
- 快速上手 · 三个典型场景
- 与同类工具的差异
- 项目结构
- 设计哲学
- 引用本工作
- 贡献
- 关于作者
- License
- 致谢
定位
人文学者的端到端写作助手——覆盖一篇人文论文从研究问题到投稿披露的完整生命周期:
研究问题 → 文献地图 → 规划 → 起草 → 修订 →
对抗性审稿 → AI 痕迹清理 → 盲读核对 → AI 使用披露
服务于**"文字本身就是论证"**的领域——历史、哲学、文学、文化研究、艺术史、宗教学、古典学、思想史、科学史以及相邻的人文导向学科。
本 skill 不是润色工具,不是引用管理器,也不是研究流水线。本 skill 是一个陪你走完整个写作弧线的思维伙伴。
覆盖写作全生命周期的 12 个模式
| 阶段 | 模式 |
|---|---|
| 写作前 | Mode H · 研究问题 sharpening · Mode I · 文献脉络梳理 · Mode J · 规划专项模式 |
| 起草 | Mode C · 构思 → 新内容写作 · Mode A · 段落对话 |
| 审读 | Mode B · 章节审读(四层批判) · Mode D · 魔鬼代言人(1-5 级 calibration + 方法论专项) |
| 修订 | Mode E · 写作瓶颈辅助 · Mode F · 底稿修订(含 revision-coach 子模式) |
| 投稿前 | Mode G · 盲读核对 · Mode K · AI 使用披露 |
| 评审后 | Mode L · 修订工作流(答辩/外审意见整合,修订档案制) |
外加引用工具链(scripts/):格式一致性扫描、格式转换(Chicago / MLA / APA / GB7714)、Crossref/OpenAlex 元数据候选检索（不核验原文或论断支持）。在具备 agent 编排能力的环境(如 Claude Code)中,Mode B/D 审读可以分派给并行审稿 agent；待核断言可查阅来源，分别记录出处、支持关系与局限。
本 skill 严肃对待的三件事
通用 AI 写作工具不会处理人文学科的这三件事：
1. 声音保持是核心学术价值，不是风格装饰。 在人文学科里，作者的声音承载认识论重量：它表明作者从哪个智识传统中写作、把谁视为对话者、哪些动作是自己的而哪些是借来的。一篇被润色成"标准学术英语"的论文，丢失的就是这种信号。本 skill 帮助你写得更像自己，而不是更不像。
2. 论证和散文是不可分离的。 实证研究里，糟糕的写作可以毁掉一个完美的实验。但人文学科里，写作就是论证本身——一句疲软的话、一个含糊的概念、一个未经辩护的过渡，都是论证失败。本 skill 在"散文中的论证"层面工作。
3. 审稿人是真实而对抗性的。 人文审稿人不温和。一个理论概念会被测试其锋利度；一个历史断言会被测试其证据；一个哲学论证会被测试最强的反对意见。本 skill 在内部模拟这种对抗，让论文在投稿前就先遇上。
一次典型的交互
你：帮我看看这段。我在论福柯的全景敞视主义如何延伸到数字平台。
[skill 读取该段 + 你的写作风格档案]
伙伴：
我注意到三件事，按优先级排序。
🔴 基底问题：你把"全景敞视"用作平台监控的隐喻，但福柯的原初概念
依赖于一种建筑学装置生产出特定的主体位置。平台生产的是另一种主体
（数据主体，而不是关在牢房里的身体）。你可以选择 (a) 把这个类比
明确为类比并讨论差异；或者 (b) 用 Zuboff 的"工具型权力"，更贴合
数据情境。你想走哪个方向？
🟡 论证问题：第二段从"平台收集数据"跳到"这构成了监控"，没有补上中间
"可见性即权力"的环节。一个善意读者会跟不上。
🟢 文风问题：你的风格档案里写"第一人称 + 破折号插入"。第三段以"本文
认为……"开头，读起来像八股。要不要恢复第一人称？
你：走方向 (b)。改吧。
[skill 给出修订；你保留听起来像你自己的版本]
这就是"思想对话伙伴，而不是润色工具"在实践中的样子。
核心特性
四层批判模式（不是单一润色）
第一层 · 基底批判 — "这篇论文在学术上成立吗？"
第二层 · 结构批判 — "论证是怎样展开的？展开得好吗？"
第三层 · 段落批判 — "这一段在做什么？做好了吗？"
第四层 · 语句批判 — "这句话说对了吗？说好了吗？"
按范围确定优先级：全文审阅先处理重要论证问题；局部修改不依赖这些问题的解决时，完成用户所要求的修改。
魔鬼代言人 + 抗谄媚机制
模拟三种审稿人 + 一个善意困惑读者：
- 审稿人 A · 理论苛刻型
- 审稿人 B · 历史实证型
- 审稿人 C · 方法论质疑型
- 读者 D · 善意困惑型（独到设计：能让善意读者困惑的地方就是论证薄弱处）
依据证据纠正：不因压力无依据让步；但一条决定性的原文或有效推理足以推翻异议，就应撤回。审稿强度不允许无视纠正。
文风深层学习与保持（"我手写我口"）
不只是句式偏好，还包括：
- 论证节奏（线性 / 螺旋 / 张弛）
- 学术姿态（批判性继承 / 对话性推进）
- 引用的修辞功能（权威锚点 / 批判靶标 / 对话接口 / 叙事性 / 概念工具）
- AI 痕迹排查清单（八类未审视表达模式 ≠ AI 套话 ≠ 学术八股）
学科特殊维度
不同章节类型对应不同的批判策略：
- 历史叙事 · 时代错位、反事实压力测试、史料处理
- 哲学论证 · 概念推演、跨理论嫁接、最强反对意见
- 文学批评 · 细读与诠释、文体意识、形式与意义
- 文化研究 · 权力-知识框架、位置性、概括范围
- 艺术史 · 描述与诠释、来源、接受史
- 宗教学 / 古典学 · 源语严谨性、传统意识、内外位置
- 案例分析（跨学科共用）
ADHD 友好的交互设计
- 反馈分批：每轮 3-5 项最多
- 快速胜利优先
- 跳跃跟随（话题跳跃可能是洞见信号）
- 番茄钟友好的任务单元
- 长对话每 4-5 轮给"我们现在在哪"摘要
自反性写作支持（独到模块）
如果作者的研究本身涉及人-AI 协作（autoethnography of AI-assisted writing），本 skill 提供六类"反思时刻"分类：
🔄 方向转变 / 🚫 拒绝 / 🎭 声音冲突 / 🔧 工具依赖 / 💡 意外洞见 / 🤖 AI 痕迹觉察
学术依据（在论文里可直接引用）：
- Christou (2026). Reconfiguring Reflexivity in the Era of AI. Qualitative Inquiry.
- Wiles (2025). Recursive Cognition in Practice. International Journal of Qualitative Methods.
- Panke (2025). How Can (A)I Research This? An Autoethnographic Exploration of Generative AI.
工程化基础设施
借鉴软件工程的最佳实践，服务于人文写作：
- 版本管理：小版本 = commit / 大版本 = release /
_drafts/= feature branch - 修改日志：每次修改记录 diff + reason（像 git commit message）
- 反馈报告：Blocker / Major / Minor / Question 四级（借鉴 code review）
- 系统性验证清单：论证完整性 / 概念一致性 / 引用完整性 / 文风一致性
[VERIFY]/[待核对]硬标记：抗引用幻觉——绝不允许进入投稿版本
十二个工作模式（不是单一对话模式）
- 模式 A — 段落级对话
- 模式 B — 章节级审读
- 模式 C — 构思 → 新内容写作（含协作式起草协议）
- 模式 D — 魔鬼代言人（1–5 级 calibration，方法论专项子模式）
- 模式 E — 写作瓶颈辅助（五种解冻策略）
- 模式 F — 底稿修订（双版本对照，对抗 AI 腔调，含 revision-coach 子模式）
- 模式 G — 盲读核对（机械检查"承诺-兑现"）
- 模式 H — 研究问题锐化（苏格拉底式对话：从模糊兴趣到可写的锋利问题，含 "so what" 测试与真实对话者识别）
- 模式 I — 文献脉络梳理（把你已经读过的文献整理成阵营与论战地图——绝不替你搜文献）
- 模式 J — 规划专项（学科感知的标准论证弧线；只做大纲，不写正文）
- 模式 K — AI 使用披露（审计实际 AI 参与度，四级分类，生成期刊可用的披露声明）
- 模式 L — 修订工作流（答辩/外审意见整合：一条意见一份修订档案，状态权威主表索引）
工程化辅助脚本
scripts/ 提供五个零依赖工具：
| 脚本 | 用途 |
|---|---|
ai-trace-scan.sh |
扫描套话与连接词堆砌 |
pending-checks.sh |
汇总所有 [VERIFY] / [待核对] / ❓ 待讨论 / [AI 草稿] 标记 |
citation-consistency.py |
引用格式一致性扫描（括号 / 逗号 / 连接词 / 中英姓名 / 页码） |
citation-format-convert.py |
BibTeX 文献表在 Chicago（著者-出版年）/ MLA 9 / APA 7 / GB/T 7714 之间转换 |
citation-verify.py |
检索 Crossref/OpenAlex 元数据候选（FOUND / FUZZY_MATCH / NOT_FOUND / ERROR）；不核验原文或论断支持 |
支持的人文学科
本 skill 用三层架构组织人文学科,以便学科路由系统真正匹配作者的工作位置——而不是套用一个扁平的七项清单。作者在 onboarding 时声明学科(或 skill 从草稿中推断),路由按对应层级加载。
L1 · 人文学科六个一级大类
这些是经典的人文学科 L1 大类。每个携带一组通用 AI 写作工具看不见的核心方法论关切。
| L1 大类 | 研究对象 | 核心方法论关切 |
|---|---|---|
| 文学 · Literature | 文本(诗、小说、戏剧、散文) | 文本细读 vs 解释 · 文类意识 · 形式-意义贴合 · 互文性 |
| 史学 · History | 过去的事件、人物、社会 | 时代错置 · 反事实压力 · 史料处理(一手 vs 二手)· 因果链透明 · 史学史定位 |
| 哲学 · Philosophy | 概念、论证、规范命题 | 概念派生 · 论证形式(形式 vs 实质)· 跨理论移植代价 · 最强反对的 steel-man · 模态范围 |
| 语言学 · Linguistics | 语言结构与使用 | 数据来源(corpus / 直觉 / 启发实验)· 形式 vs 功能 · 描述 vs 规定 · 跨语言主张范围 |
| 艺术学 · Art Studies | 艺术作品(绘画、雕塑、音乐、电影、建筑) | 描述 vs 解释(必须分开)· 来源与物质性 · 接受史 · 媒介特异的形式分析 |
| 宗教学 · Religious Studies | 宗教传统、文本、实践 | 源语言严谨度(原文 vs 译本)· 传统位置 · 内部/外部(emic vs etic)· 比较方法 |
L2 · 常见子学科(非穷举)
子学科继承父 L1 的全部方法论关切,外加作者在 onboarding 时声明的具体约束。例子——还有更多可能:
| 父 L1 | 子学科示例 |
|---|---|
| 文学 | 中国古代文学 · 中国现当代文学 · 比较文学 · 文学理论 · 文学批评 · 外国文学 |
| 史学 | 中国史 · 世界史 · 经济史 · 社会史 · 文化史 · 城市史 · 断代史(唐史、近代史等) |
| 哲学 | 中国哲学 · 西方哲学(分析 / 大陆)· 伦理学 · 美学 · 政治哲学 · 科学哲学 · 现象学 |
| 语言学 | 历史语言学 · 社会语言学 · 语用学 · 类型学 · 话语分析 |
| 艺术学 | 艺术史 · 音乐学 · 电影学 · 戏剧学 · 建筑史 |
| 宗教学 | 基督教研究 · 佛教研究 · 道教研究 · 宗教比较学 |
如果你的子学科未列出,在 onboarding 时声明即可——自动继承父 L1。
L3 · 跨学科 / 交叉领域(显式多重继承)
这些是人文领域中显式从多个 L1 取材的领域。skill 会加载所有父 L1 的方法论关切 + L3 特化叠加。
| L3 领域 | 继承自 | L3 特化叠加 |
|---|---|---|
| 文化研究 · Cultural Studies | 文学 + 史学 + 社会学 | 权力-知识框架 · 位置性 · 概括范围 |
| 古典学 · Classics | 文学 + 史学 + 哲学 + 宗教学 + 考古学 | 文本批评(写本传统)· 语文学严谨度 · 接受史 |
| 思想史 · Intellectual History | 史学 + 哲学 | 概念史方法(Begriffsgeschichte vs Cambridge School)· 语境 vs 文本 · 避免现时主义 |
| 科学史 · History of Science | 史学 + 科学 + 哲学 | 内史 vs 外史 · 辉格史警惕 · 技术准确性 · 案例研究校准 |
| 媒介研究 · Media Studies | 文学 + 文化研究 + 技术哲学 | 媒介形态学 · 接受研究 · 技术-社会共构 |
| 数字人文 · Digital Humanities | 任一 L1 + 计算 | 数据可重复性 · 工具透明 · 算法偏倚 · 计算选择的方法论披露 |
| 性别研究 · Gender Studies | 文学 + 史学 + 文化研究 | 性别本体论 · 历史化性别 · 交叉性 |
| 后殖民研究 · Postcolonial Studies | 文学 + 史学 + 文化研究 | 位置性 · 翻译政治 · 抵抗欧洲中心主义 |
| 环境人文 · Environmental Humanities | 文学 + 史学 + 科学 | 人类纪框架 · 多物种视角 · 尺度问题(局部 vs 行星) |
人文邻近领域(欢迎,带 scope 注释)
有些领域形式上归在社会科学,但包含强烈人文导向的子传统(文字本身就是论证)。本 skill 欢迎这类工作:
| 领域 | 服务什么 | 不服务什么 |
|---|---|---|
| 传播学 · Communication Studies(人文路径) | 媒介环境学派(Innis / McLuhan / Postman / Carey);批判传播学;人文导向的文化-媒介研究 | 实证 / 实验传播学,作为量化方法的内容分析 |
| 教育学 · Educational Research(人文路径) | 教育史 · 教育哲学 · 人文模式的课程理论 · 批判教育学 | 量化教育心理学 · 实证学习成效研究 |
继承关系:传播学(人文路径)← 媒介研究 + 哲学 + 文化研究。教育学(人文路径)← 史学 + 哲学 + 文化研究。
兜底协议——如果你仍然找不到自己的领域
在 onboarding 时告诉 skill:
- 研究对象:文本 / 过去事件 / 概念 / 现象 / 物件 / 实践 / 语言结构 / ……
- 主要方法:文本细读 / 档案研究 / 论证分析 / 民族志 / 比较 / 语文学工作 / ……
skill 会推断最接近的 L1 + 相关 L3 叠加,写入 _writing-config/学科档案.md,然后继续。你可以随时细化声明。
每个 L1 / L3 entry 的完整方法论 rubric 在 references/disciplines.zh.md 中——本 README 表是表面地图,skill 内部承载完整 rubric。
Showcase · 修改前后示例
这是教学练习，不是某位学者的已核验引语。刻意只提供有限材料：
原稿。 所有学徒主要依靠书面规程学习。两次工坊访谈证明，正式规则总是优先于经验。 作者提供的阅读笔记。 两次访谈中，学徒都说常规操作时查阅书面规程，遇到例外情况时请教有经验的工人。
skill 应在强化文字前指出材料与主张的距离：
发现 · 主张与证据不匹配。 访谈笔记描述了两种学习方式，没有证明规则总是优先。两次工坊访谈也不能支撑“所有学徒”。该判断依据提供的笔记，尚未核对访谈原始记录。
如果作者要求修改，可提出范围有限的版本：
修改后。 这里概述的两次访谈中，学徒描述了常规操作时查阅书面规程、遇到例外时请教有经验工人的做法。这些叙述提示，规则与经验的作用可能随任务而变化。[待核对：查访谈原始记录并补充真实定位。]
修改让论断范围与已有材料相称，并保留来源核验缺口。不编造引语、页码或外部学术权威；这项解释是否符合全部证据，仍由作者判断。
安装
作为 Claude Code skill 安装
git clone https://github.com/tizzy916/humanities-writing-companion.git \
~/.claude/skills/humanities-writing-companion
chmod +x ~/.claude/skills/humanities-writing-companion/scripts/*.sh
或者作为项目级 skill（仅当前 vault / project 可用）：
git clone https://github.com/tizzy916/humanities-writing-companion.git \
./.claude/skills/humanities-writing-companion
Claude Code 加载
Claude Code 启动时会自动扫描 ~/.claude/skills/ 和 ./.claude/skills/。安装后说"我在写人文论文"或下方任意触发词即可激活。
不用 git（ZIP 下载）
无需 git：在 GitHub 仓库页面 点击 Code → Download ZIP，解压后把解压出的文件夹移动到 ~/.claude/skills/humanities-writing-companion（项目级安装则放到 ./.claude/skills/humanities-writing-companion），然后给 shell 脚本加执行权限：
chmod +x ~/.claude/skills/humanities-writing-companion/scripts/*.sh
Claude 桌面版 / claude.ai
Claude 桌面版与 claude.ai 也支持自定义 skill：把 skill 文件夹（即包含 SKILL.md 的目录）打包为 .zip，在 Claude 设置中的能力/skills 相关板块上传（具体菜单措辞可能随产品迭代变化——找 "Skills" 字样即可）。注意：scripts/ 工具链需要可执行 shell 的环境（Claude Code / agent 模式）；纯对话环境下 skill 的各对话模式可用，但脚本不会运行。
Claude Agent SDK 接入
SKILL.md 可直接加载到系统提示词中。skill 是纯文本，无运行时依赖。
其他 agent（开放 SKILL.md 格式）
本 skill 采用开放的 Agent Skills 格式：一个包含 SKILL.md 以及纯文本 references/、scripts/ 的文件夹。任何支持该格式的 agent——或者只要能把 SKILL.md（及按需路由到的 references/*.md）读入上下文——都可以使用：把本仓库 clone 到你的 agent 发现 skills 的目录即可。scripts/ 工具链需要 zsh 和 Python 3。引用查询还需要联网访问 Crossref 和 OpenAlex；其余检查可在本地运行。
验证安装
安装后新开一个对话，二选一：
- 直接问 Claude："你现在加载了哪些 skills？"——列表中应出现
humanities-writing-companion；或 - 直接说一个触发词，例如 "帮我看看这段" 或 "review my section"——skill 应当激活，并以四层批判的方式回应，而不是像通用润色工具那样回答。
如果两者都不生效，检查文件夹是否直接位于 ~/.claude/skills/ 之下（即 ~/.claude/skills/humanities-writing-companion/SKILL.md 存在），然后重启 Claude Code。
触发词
中文：论文 · 写作 · 润色 · 改论文 · 帮我看看这一章 · 我手写我口 · 这个论证有没有问题 · 我写不下去了 · 审稿人会怎么攻击
英文："paper," "essay," "chapter," "dissertation," "argument," "thesis," "revise," "voice," "review my section," "stuck on writing," "devil's advocate"
已有学术稿件语境时，随口说“帮我看看这段话”或“take a look at this paragraph”也适用。
快速上手 · 三个典型场景
场景 1：新论文初次使用
对 Claude 说："我想写一篇关于 X 的论文。"
skill 从你的想法和已有材料出发，帮助形成研究问题或交付所需草稿，只询问会影响结果的缺失信息。选择持续文件协作后，再按需建立档案与目录。
场景 2：修改已有章节
"帮我看看这一章" → 模式 B（章节级审读）→ 四级反馈报告
"帮我改这段" → 模式 A（段落级对话）→ 诊断 + 建议 + 理由
"我写不下去了" → 模式 E（写作瓶颈）→ 一个合适的下一步
场景 3：对抗 AI 腔调（双版本对照）
如果你的论文经过 AI 润色，但想恢复原始文风：
模式 F · 底稿修订 → 对照 AI 润色版与原始版本 → 保留改善 + 恢复声音
与同类工具的差异
| 工具 | 它的定位 | 本 skill 与之差异 |
|---|---|---|
| Jenni AI | 实时自动续写 + 文献发现 | 我们做思想对话,不做续写。实时预测会跳过人文论证所需的认知工作 |
| Paperpal | 学术语言润色(偏理科/生医) | 我们是写作架构(12 个模式 + 四层批判 + 学科路由),不是单点润色工具 |
| Yomu AI | Sourcely 文献引擎 + 段落反馈 | 文献由作者自管(Zotero / Drive)。Mode I 帮你整理读过的,从不替你读你没读的 |
| Thesify | Paper Digest + Purpose-Check | Mode G 借鉴了 Purpose-Check,但放在四层批判 + reviewer calibration 的更大工作流里 |
| HyperWrite Devil's Advocate | 单点反方论证生成 | Mode D 是完整模式:1-5 级 calibration + 方法论专项 + 基于证据的让步机制(抗谄媚) |
| Grammarly / DeepL Write | 语法 / 翻译润色 | 我们绝不为了"清晰"牺牲声音。「我手写我口」是核心原则不是可选项 |
| 通用 ChatGPT / Claude(无 skill) | 通用对话 | 我们跨对话持续维护:写作风格档案、读者档案、修改日志、四层批判、学科路由、AI 痕迹清单、引用工具链 |
项目结构
humanities-writing-companion/
├── SKILL.md ← 核心 skill 文件(英文,精简入口:原则、路由表、四层批判、模式存根)
├── SKILL.zh.md ← 中文镜像版
├── references/ ← 按需加载手册(每个都有 `.zh.md` 中文镜像)
│ ├── critique-review.md ← 详细四层问题与审阅覆盖检查
│ ├── disciplines.md ← 完整学科维度表(L1/L2/L3/邻近 + 兜底协议)
│ ├── modes-prewriting.md ← 模式 H / I / J 完整协议
│ ├── mode-c-drafting.md ← 模式 C 四阶段起草流程
│ ├── mode-d-adversarial.md ← 模式 D 魔鬼代言人完整协议
│ ├── mode-e-bottleneck.md ← 模式 E 写作瓶颈策略
│ ├── mode-f-revision.md ← 模式 F 底稿修订工作流
│ ├── modes-submission.md ← 模式 G / K 完整协议
│ ├── deep-style.md ← 文风深层理解与保持
│ ├── multilingual-writing.md ← 多语言写作规范
│ ├── style-profile-template.md ← 写作风格档案("声音宪法")模板
│ ├── ai-trace-checklist.md ← AI 痕迹排查清单
│ ├── project-management.md ← 项目文件夹 + 版本管理规范
│ ├── revision-workflow.md ← Mode L 修订档案制工作流手册
│ └── target-reader-profile-template.md ← 目标读者档案模板
├── scripts/ ← 工程工具链(零依赖)
│ ├── README.md ← 脚本使用说明
│ ├── ai-trace-scan.sh ← AI 套话扫描(zsh)
│ ├── pending-checks.sh ← 待办标记汇总(zsh)
│ ├── citation-consistency.py ← 引用格式一致性(Python 3)
│ ├── citation-format-convert.py ← Chicago/MLA/APA/GB7714 转换(v4.0+)
│ └── citation-verify.py ← Crossref/OpenAlex 元数据候选检索
├── README.md ← 英文 README
├── README.zh.md ← 本文件
├── CHANGELOG.md ← 版本历史
├── LICENSE ← CC BY-NC 4.0
└── CITATION.cff ← 学术引用元数据
双语状态：项目已完全双语化。SKILL.md、README、CONTRIBUTING、references/ 的全部手册以及 scripts/README 均为英文文件 + .zh.md 中文镜像成对存在；脚本注释同样双语。两种语言的触发词都能激活 skill（SKILL.md 的 description 字段同时处理两种语言）。
设计哲学
"我手写我口"
学术严谨与个人表达不对立。"标准学术语体"往往意味着个性的消亡。skill 帮助作者用自己的声音说话，而不是把文字压入预制模具。
思想优先，格式其次
修改优先级：
- 论证的力量
- 概念的精确
- 结构的有效
- 表达的质量
- 格式的规范
永远从上往下工作。不要在一个论证有根本缺陷的段落里纠结逗号。
工程化严谨，人文化表达
借鉴软件工程的最佳实践（版本管理、单元测试、code review），但服务于人文写作的特殊性。工程化不是把论文变成代码，而是让每次修改可追溯、论证质量可验证、写作过程可接续、问题分层处理。
引用本工作
如果你的研究使用了本 skill，请在方法论部分引用。
BibTeX:
@software{shen_humanities_writing_companion_2026,
author = {Shen, Cong},
title = {Humanities Writing Companion: An Agent Skill for Voice-Preserving Humanities Academic Writing},
year = {2026},
publisher = {Zenodo},
version = {5.1.0},
doi = {10.5281/zenodo.20280772},
url = {https://doi.org/10.5281/zenodo.20280772}
}
纯文本署名(用于 skill 元数据、页脚等):
Based on Humanities Writing Companion by Shen Cong
https://github.com/tizzy916/humanities-writing-companion
完整机读元数据见 CITATION.cff(GitHub 的 "Cite this repository" 按钮会自动调用该文件)。
同时引用 Companion 工具
如果你在同一项目中同时使用了 academic-research-skills,请同时引用两者。ARS 的署名格式(遵循 CC BY-NC 4.0):
Based on Academic Research Skills by Cheng-I Wu
https://github.com/Imbad0202/academic-research-skills
贡献
欢迎 issue 与 PR：
- 新的工作模式提议
- AI 痕迹排查清单的扩充
- 学科特异性示例（中世纪研究、艺术保护、民族音乐学等）
- 不同引用格式（APA / Chicago / MLA / GB/T 7714 / 期刊自定义）的支持
- 双语镜像的翻译质量改进（
references/、scripts/）
关于作者 / About the Author
沈聪,中央美术学院实验艺术学院本科,清华大学科学史系硕士(导师 胡翌霖),科技文创公司 天与视界 TIANYU VISION 创始人 & CEO。
本 skill 诞生于学位论文《技术自由主义》的写作过程——作者发现市面上的 AI 写作工具几乎都偏向"润色与平均化",而人文学术写作真正需要的是反向的能力:保护作者的学术声音、检验论证的严密性、扛住真实审稿人的对抗。所以做了这个 skill——不是替自己写,而是替自己读,在四个层级(基础严密性 / 论证展开 / 段落功能 / 句子措辞)分别提供一个真实人文学者会给出的批评。
📮 GitHub @tizzy916 · shencong916@gmail.com · 论文、合作、纠错欢迎来信
License
CC BY-NC 4.0(知识共享 署名-非商业性使用 4.0 国际许可协议)——非商业用途自由使用、修改、分发,要求 attribution。
⚠️ License change (v3.0.0, 2026-05-19):本项目从 MIT 改为 CC BY-NC 4.0。v2.1.0 及更早版本仍按 MIT 发布,保留原始商用权利(仅限这些特定版本)。从 v3.0.0 起,未经单独授权禁止商业用途。
商业用途 / Commercial Use
本 skill 采用 CC BY-NC 4.0 协议——仅限非商业用途(学术研究、教学、个人项目、开源衍生、机构内部研究流程)。
如需商业使用——嵌入付费产品、使用本 skill 提供付费咨询或编辑服务、商业 SaaS 集成、代客商业写作服务——请联系作者获取商业 license:
📮 shencong916@gmail.com(沈聪 · 天与视界 TIANYU VISION)
作者保留按个案授予商业 license 的权利。在学术发表中引用本 skill 不受 license 层级影响,始终允许。
致谢
本 skill 的方法论灵感与学术依据：
- Christou, P. A. (2026). Reconfiguring Reflexivity in the Era of AI. Qualitative Inquiry.
- Wiles, F. (2025). Recursive Cognition in Practice. International Journal of Qualitative Methods.
- Panke, S. (2025). How Can (A)I Research This?.
- Foucault, M. (1984). What is Enlightenment? — "对当下的诊断"作为方法论传统
- Stiegler, B. (2013). What Makes Life Worth Living: On Pharmacology — "批判药理学"
部分设计模式参考：
- Imbad0202/academic-research-skills — 本 skill 作为互补的上游流水线；他们的 reviewer 模块中的 Concession Threshold 模式启发了模式 D 的"让步前最低标准"
- Voice DNA + Audience Profile 模式 — 启发了写作风格档案与目标读者档案的配对设计
- Thesify Purpose-Check — 启发了模式 G 盲读核对