v5.0.2 — Fix Concept DOI (20280773 → 20280772)

Citation-metadata fix only — no skill content changed.

• Concept DOI corrected everywhere: all "Concept DOI" references pointed to 10.5281/zenodo.20280773, which is actually the v4.1.1 version-specific DOI — it permanently resolves to the v4.1.1 archive (old title, old repository name) instead of the latest version. The true Concept DOI is 10.5281/zenodo.20280772, which always resolves to the newest archived release. Corrected in the README DOI badges and BibTeX blocks (both languages), CITATION.cff identifiers, the release checklists, and the v4.1.1 changelog entry's DOI link.

Full details in CHANGELOG.md.

v5.0.1 — Platform-neutral repositioning: renamed to humanities-writing-companion

Docs and metadata only — no skill content changed.

• Repository renamed: claude-skill-humanities-writing-companion → humanities-writing-companion (GitHub redirects all old URLs). Every hardcoded repository URL updated: READMEs, CITATION.cff, .zenodo.json, docs/, issue-template config, and the citation-verify.py User-Agent.
• Positioning de-branded from Claude: the skill is written in the open Agent Skills (SKILL.md) format and is not Claude-only. Updated accordingly: README taglines and format badge, citation titles ("A Claude Skill…" → "An Agent Skill…") in CITATION.cff / .zenodo.json / README BibTeX, the claude-skill keyword (→ agent-skill), the GitHub repo description, and CONTRIBUTING's entry-point list. New "Other agents (open SKILL.md format)" install note in both READMEs. Factual Claude Code / claude.ai / Agent SDK install instructions are retained.

Full details in CHANGELOG.md.

v5.0.0 — Audit-driven fixes + progressive-disclosure restructure

Ships v4.3.0 (audit-driven fixes) and v5.0.0 (restructure) together. Full details in CHANGELOG.md.

v5.0.0 · Progressive-disclosure restructure 渐进式披露重构

• SKILL.md body ~2,000 → ~840 lines — per-activation context cost drops from ~35k to ~18k tokens. 主文件瘦身一半以上，每次激活的上下文成本减半
• 9 new bilingual reference pairs, loaded on demand: disciplines · modes-prewriting (H/I/J) · mode-c-drafting · mode-d-adversarial · mode-e-bottleneck · mode-f-revision · modes-submission (G/K) · deep-style · multilingual-writing. Sections moved verbatim; every mode keeps a decision-carrying stub in the core file. 九对按需加载的双语手册；核心文件中每个模式保留带硬约束的存根
• Selective Loading Guide rewritten as the actual router (task → which file to Read). 按需加载指南重写为真正的路由表

v4.3.0 · Audit-driven fixes 审计驱动修复

A 31-agent QA audit (14 mode simulations, 4 static audits, 7-channel market scan, 5-judge panel) → every confirmed defect fixed:

• Routing: frontmatter description 1,727 → 992 chars (was 69% over the 1,024-char Agent Skills spec limit — the most aggressive triggers never reached the router on truncating platforms) 触发描述压缩至规范限内
• Toolchain integrity: citation-verify.py no longer rubber-stamps hallucinated citations (FOUND now prints the matched title for human confirmation; new ERROR verdict; Crossref→OpenAlex cascade); ai-trace-scan.sh directory mode actually works now (previously printed a false "✅ no AI traces") 反幻觉脚本的两处目的反转缺陷修复
• Mode upgrades: Mode E first-response protocol + bottleneck typology + capability boundary; Mode L four-way triage + response-letter/修改说明 workflow + traceability matrix; Mode K policy-never-from-memory + real Template C + anti-whitewashing rules; Mode D evidence contract + review-the-review + two-stage option
• New: references/style-profile-template bilingual pair (the "constitution of voice" finally has a template); rule-precedence + global flagged-diff rules; minimal-start protocol; 58-assertion test suite + CI (bilingual heading-parity across 18 file pairs); release checklist
• 12 mechanisms adapted from the strongest peer tools (evidence contract, review-the-review guardrail, rhetorical-action menu, process-ledger disclosure, over-imitation reverse check, …)

Compatibility 兼容性: existing _writing-config/ / _meta/ project files continue to work unchanged. The skill has required multi-file support since v4.x; packaging shape is unchanged. 既有项目文件无需迁移。

v4.1.1 · First Zenodo-archived release · DOI minting trigger

First Zenodo-archived release of this repository.

Functionally identical to v4.1.0. Version bump only. The purpose of this release is to trigger the first Zenodo archive event for this repository, establishing a permanent DOI that can be cited in academic work.

What this means for citation

Once Zenodo finishes archiving (typically 1–2 minutes after this release publishes), this repository will have a permanent DOI of the form 10.5281/zenodo.NNNNNNN. The DOI will be:

• Version-specific: each future release gets its own DOI (this one for v4.1.1)
• Concept-DOI: a version-agnostic DOI that always resolves to the latest release

Papers using this skill can then cite:

Shen, C. (2026). Humanities Writing Companion: A Claude Skill for
Voice-Preserving Humanities Academic Writing (Version 4.1.1) [Software].
Zenodo. https://doi.org/10.5281/zenodo.NNNNNNN

CITATION.cff was already in place since v2.1.0; .zenodo.json since the initial Zenodo setup commit. This release just bumps the CITATION.cff version field to match the release tag for Zenodo's archival metadata.

See v4.1.0 release notes for the substantive changes (3-layer discipline architecture). No functional / behavioral changes in this release.

v4.1.0 · Discipline architecture: flat 7-entry → 3-layer

Discipline architecture refactor.

The previous "Supported humanities disciplines" section listed seven entries as if they were at the same conceptual layer. They were not — three were L1 main disciplines, one was an L2 subfield, two were L3 cross-disciplinary fields. The set also reflected the author's own research range rather than the actual map of humanities scholarship.

This release restructures the discipline architecture so users from any humanities or humanities-adjacent field can find their position.

Three-layer architecture

L1 · Six main humanities disciplines

Literature · History · Philosophy · Linguistics (new) · Art studies (broader parent of the prior "Art history") · Religious studies

Each gets a refined methodology rubric with 5-7 core concerns.

L2 · Subfield inheritance

Subfields (e.g., 中国古代文学, 经济史, 伦理学, 艺术史, 音乐学, 历史语言学) inherit all parent L1 concerns and may add subfield-specific overlays. Documented common subfield overlays for each L1.

L3 · Nine cross-disciplinary fields with explicit multi-inheritance

L3 field	Inherits from
Cultural studies	Literature + History + Sociology
Classics	Literature + History + Philosophy + Religious studies + Archaeology
Intellectual history (new)	History + Philosophy
History of science (new)	History + Science + Philosophy
Media studies (new)	Literature + Cultural studies + Philosophy of technology
Digital humanities (new)	Any L1 + Computation
Gender studies (new)	Literature + History + Cultural studies
Postcolonial studies (new)	Literature + History + Cultural studies
Environmental humanities (new)	Literature + History + Science

Humanities-adjacent fields welcomed

• Communication studies (humanities-style) — Media ecology school (Innis / McLuhan / Postman / Carey); critical communication; rhetorical studies; philosophy of communication
• Educational research (humanities-style) — History of education; philosophy of education; critical pedagogy

Each with explicit "what we serve / what we don't serve" boundaries.

Fallback protocol

For any field not on the list: author declares object of study + primary method, skill infers closest L1 + relevant overlays.

Implementation

• Onboarding step 1 upgraded to 3-layer elicitation (L1 required, L2/L3 optional)
• _writing-config/discipline.md schema updated (backward compatible)
• Mode D methodology-focus attack table restructured: 6 L1 vectors + 9 L3 overlays + 2 humanities-adjacent
• Tag format: [Mode D · methodology · L1 X] / [Mode D · methodology · L3 Y]
• Mode J standard arc table restructured: 6 L1 arcs + selected L3 / adjacent arcs (incl. media-ecology arc, history-of-education arc)

Migration

Backward compatible: v4.0.0 discipline.md files still parse correctly. New users in communication studies / educational research who previously had no entry point should re-declare in onboarding using the new humanities-adjacent option.

See CHANGELOG.md for full history.

v3.0.0 · License change to CC BY-NC 4.0

⚠️ BREAKING CHANGE — license switch 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 onward, commercial use is prohibited without a separate license.

Why

• Aligns with academic-research-skills (the companion empirical-research pipeline, then a designated companion), which uses the same license
• Reserves commercial rights for the author's own work (Tianyu Vision)
• Permits all non-commercial use: academic research, teaching, personal projects, open-source derivatives, internal research workflows

What changed

• LICENSE: replaced MIT text with full CC BY-NC 4.0 legal code from creativecommons.org
• README.md / README.zh.md: badge swapped; License section rewritten with change notice + Commercial Use subsection
• CITATION.cff: license field MIT → CC-BY-NC-4.0; version 2.1.0 → 3.0.0
• CONTRIBUTING.md: contributor agreement updated to CC BY-NC 4.0
• .zenodo.json: license MIT → CC-BY-NC-4.0; creator name properly upgraded to "Shen, Cong"
• BibTeX version bumped to 3.0.0

Commercial licensing

For inquiries: shencong916@gmail.com (Shen Cong · Tianyu Vision).

See CHANGELOG.md for full history.

v2.1.0 · ARS companion + citation conventions + discipline routing

Compliance + humanities-scholar focus pass.

Added

• Dedicated ## Companion: academic-research-skills section with division-of-labor diagram and design-lineage attribution (Concession Threshold pattern)
• ## Supported humanities disciplines README section: seven main disciplines + cross-disciplinary case, each with failure modes the skill watches for
• ## Showcase · Before / After example demonstrating four-layer critique refusing to descend below a broken foundation
• ### Discipline routing protocol in SKILL.md: discipline becomes a load-bearing routing variable
• Citation section enhanced with proper BibTeX, plain-text attribution block, and Citing-companion-tools subsection per CC BY-NC 4.0
• CITATION.cff: full author identification (Shen Cong + alias + affiliation + website); ARS added as type: software reference

Changed

• Onboarding step 1 upgraded to enforce explicit discipline elicitation with seven canonical options

License

Still MIT in this release (pre-v3.0.0 relicense — see v3.0.0 release notes).

See CHANGELOG.md for full history.

v2.0.0 · Humanities pivot + bilingual rewrite

Major repositioning + bilingual support.

Changed

• Repository renamed: claude-skill-academic-writer → claude-skill-humanities-writing-companion
• Scope narrowed from broad academic writing to humanities-only (history, philosophy, literature, cultural studies, art history, religious studies, classics)
• Added bilingual SKILL.md (EN) + SKILL.zh.md (CN), plus README.md + README.zh.md
• Expanded discipline-specific dimensions to 7 areas

License

MIT (pre-v3.0.0 relicense — see v3.0.0 release notes).

See CHANGELOG.md for full history.

v4.0.0 · End-to-end humanities writing assistant

Major strategic repositioning + capability expansion.

This release removes the "humanities-side companion to academic-research-skills" framing. v4.0 is an independent end-to-end writing assistant for humanities scholars covering the full lifecycle of a humanities paper, from research-question sharpening to AI-use disclosure for submission.

Four new modes (pre-writing + submission-end coverage)

• Mode H · Research-question sharpening (Socratic) — humanities-specific question taxonomy (re-reading / re-construction / intervention), so-what test, interlocutor identification. Output: _writing-config/research-question.md.
• Mode I · Literature mapping — organizes references the author has already read into a camp-and-debate map. Hard rule: does not search literature for the author. Optional scholar-wendao integration. Output: _writing-config/literature-map.md.
• Mode J · Plan-only outlining — discipline-aware, pulls discipline-specific standard arcs. Hard constraint: does NOT write paragraphs. Output: _writing-config/outline.md.
• Mode K · AI-use disclosure (humanities-journal-specific) — 4-tier scheme + 3 template options. Audits interaction-log + revision-log. Warns when tier exceeds typical humanities journal policy. Output: _meta/AI-use-statement.md.

Enhanced existing modes

• Mode D calibration: 1-5 reviewer intensity levels (gentle reader → adversarial committee member). Concession Threshold tightens at Level 5.
• Mode D methodology-focus sub-mode: attacks methodology rather than content. Discipline-aware vulnerability set.
• Mode F.coach sub-mode: skill withholds revision until author answers 3-5 diagnostic questions. Trains author to internalize four-layer critique.

Citation toolchain (new)

• scripts/citation-format-convert.py: BibTeX → Chicago / MLA / APA / GB/T 7714 converter. Zero deps. Sanity-tested in all four styles.
• scripts/citation-verify.py: in-prose citations verified against Crossref API. Three verdicts (FOUND / FUZZY_MATCH / NOT_FOUND). Rate-limited 1 req/sec.

Removed

• ## Companion: academic-research-skills section. ARS demoted from designated companion to one of several lateral tools.
• ARS row removed from Comparison-with-adjacent-tools table.

See CHANGELOG.md for full history.

v1.0.0 · initial release

Initial public release of academic-writer skill.

Highlights

• 7 work modes (paragraph dialogue / chapter review / new-content drafting / devil's advocate / writer's-block / draft revision / blind-read check)
• 4-layer critique (foundation / structure / paragraph / sentence)
• Anti-sycophancy concession threshold for devil's advocate
• Voice preservation with style-profile + audience-profile pairing
• Reflexive-writing support with 6 reflexive moments and academic grounding
• ADHD-friendly interaction (3-5 items per round, quick wins first)
• Engineering infrastructure: version control metaphor, AI-trace scanning, pending-checks aggregation, citation consistency

Files

• SKILL.md (~1100 lines)
• references/ — AI-trace checklist, project-management spec, target-reader template
• scripts/ — 3 zero-dependency tools (zsh + Python 3 stdlib)