Roadmap

🌐 English · 中文 · 🏠 Home

Roadmap

This page has two clearly separated parts:

1. Released milestones — what has actually shipped, distilled from the CHANGELOG.
2. Directions ahead — where the project could grow. These are community-driven directions, not dated commitments.

The honest through-line: this project started life as a polishing companion and has become an end-to-end thinking partner for humanities writing — covering the whole lifecycle from sharpening a research question to writing the AI-use disclosure for submission. That shift is the story the milestones below tell.

---

Released milestones

A "polishing companion" → "end-to-end thinking partner" timeline. Every version below is a real, tagged release.

Version	Theme	What changed
v1.0.0	First public release	Shipped as academic-writer — the original polishing-companion skill.
v2.0.0	Bilingual rewrite + rename	Bilingual SKILL.md / README (English + Chinese). Repository renamed to claude-skill-humanities-writing-companion.
v2.1.0	Discipline routing + ARS companion	Discipline becomes a load-bearing routing variable; onboarding enforces explicit discipline elicitation. Added the academic-research-skills (ARS) companion section, citation conventions, and a Before/After showcase.
v3.0.0	License change	MIT → CC BY-NC 4.0. Versions ≤ v2.1.0 stay MIT; from v3.0.0 on, commercial use requires a separate license. Author identification upgraded to Shen, Cong with affiliation.
v4.0.0	Repositioned as an end-to-end assistant	The strategic turn. No longer "the writing half of an ARS pairing" — now an independent end-to-end writing assistant spanning the full humanities-paper lifecycle. Added Mode H (research-question sharpening), Mode I (literature mapping), Mode J (plan-only outlining), Mode K (AI-use disclosure); made Mode D calibratable (reviewer intensity levels 1–5) and added its methodology-focus sub-mode; added Mode F.coach. Added the citation scripts citation-format-convert.py and citation-verify.py.
v4.1.0	Discipline architecture refactor	Flat 7-entry discipline list → 3-layer L1/L2/L3 architecture. L1 = 6 main disciplines; L2 = subfield inheritance; L3 = 9 cross-disciplinary fields with explicit multi-inheritance. Added humanities-adjacent welcome (communication / educational research) and a fallback protocol for any field not on the list.
v4.1.1	First archived release	First Zenodo-archived release, with a citable DOI (10.5281/zenodo.20280773).

The through-line

v1.0.0          v2.x                v3.0.0           v4.0.0                    v4.1.x
academic-       bilingual +         relicense        end-to-end                3-layer disciplines
writer          rename +            (CC BY-NC 4.0)   assistant                 + first archived
(polishing      discipline                           (modes H/I/J/K,           release (DOI)
 companion)     routing                              calibratable D)
   │               │                   │                │                         │
   └───────────────┴───────────────────┴────────────────┴─────────────────────────┘
        from "polishing companion"  ────────────▶  to "end-to-end thinking partner"

The widening scope is deliberate. v1–v2 sharpened what the skill is for (humanities, thought-first, voice-preserving). v3 settled how it's licensed (an open, non-commercial public good). v4 widened what it covers — from a tool you reach for at the polishing stage to infrastructure you work on top of across an entire project, from the first vague topic to the submission-ready disclosure statement.

After v4.1.1 — bilingual / internationalization work

Since the first archived release, the visible work has been about making the project genuinely usable in two languages and approachable to contributors from anywhere:

• An English CONTRIBUTING.md (alongside the existing Chinese one).
• Bilingual GitHub issue templates.
• Full documentation translation — tier-1 and tier-2 docs now ship as paired .md / .zh.md files (including the design-philosophy and cross-domain-testing docs).
• Bilingual citation-script comments (中英双语) across the scripts/ toolchain.
• This Wiki itself, as a bilingual entry point.

---

Directions ahead

These are community-driven directions, not dated commitments. There are no promised dates and no promised version numbers below. This section describes where the project could grow — and is an open invitation to help it grow. Each item below maps to something the project actively values; see Contributing for the full picture and the submission process.

Discipline testing & new overlays (most valued)

The skill grew out of one specific humanities dissertation project, but its goal is to cover the entire humanities. The single most valuable contribution is to use it for real in your own discipline, report which modules fit and which don't, and propose discipline-specific overlays — for example Latin verification for medieval studies, material-culture analysis for art conservation, or field-note handling for ethnomusicology. Generality, in this project's view, should arise from the intersection of many particulars, not from a hollow abstraction. The design-philosophy doc puts the same invitation plainly: authors from other disciplines are invited to test the skill precisely so that the design can grow up amid more particulars.

Expanding the AI-trace checklist

The references/ai-trace-checklist.md is the skill's "defense in depth." There is room to grow the catalogue of unexamined expression patterns — not just AI clichés, but discipline-specific formulae and the stylistic inertia that builds up from heavy theory reading. New patterns that don't fit the existing categories (and recur often) are exactly what's wanted.

Bilingual completion of remaining files

SKILL.md and the README are already bilingual, but some supporting files under references/ remain mostly in Chinese. Completing the English side — including finding the equivalent English filler phrases for the Chinese cliché list (not literal translations) — remains a high-value direction.

New working modes

The skill currently has 11 working modes (A–K). A new mode is possible, but the bar is deliberately high — a proposed mode must:

• be mechanically distinct from existing modes (not a rename or restatement),
• have independent input requirements and output format, and
• come with at least one concrete use case plus its expected effect.

(Large additions like a new mode should start with an issue to discuss, because SKILL.md is tightly interlocking by design.)

Deeper citation-style support

Onboarding lets the author pick a style (APA / Chicago / MLA / GB/T 7714 / journal-specific). There's room for deeper built-in support for a given standard — for instance, auto-generating a reference list that complies with GB/T 7714's numeric sequential system.

Script tooling

The scripts/ toolchain can grow with new low-dependency tools — candidate ideas include paragraph-coherence detection, concept-drift detection (tracking a term's contextual frequency across chapters), and citation-density analysis (flagging anomalous passages). New scripts must stay zero- or low-dependency (zsh / Python 3 standard library / at most one or two common packages) and follow the design principles in scripts/README.md.

---

A note on what this project is not growing toward

To keep the directions above honest, it's worth stating what's deliberately out of scope — these aren't "bad" ideas, they just diverge from the design stance:

• Making the skill more general-purpose (a generic polishing mode, or extending into empirical social science / STEM). The skill is intentionally opinionated: humanities, thought-first, voice-preserving.
• Replacing modules with runtime LLM calls — the modes are plain-text prompts so they work at any Claude entry point.
• "AI smart-polishing" features — the skill's core premise is a rejection of the auto-polish narrative.
• Paid / subscription integrations — the skill is an open-source public good.

If you'd like to push on any of the directions above (or argue that one of the non-goals is wrong), that conversation is welcome — see Contributing.