antiqiusama/antiqiu-skills

Improve the clarity and voice of AI-assisted academic writing (papers, theses, rebuttals) and funding proposals (NSF Project Summary/Description, NIH Specific Aims): preserve scholarly conventions, match claims to evidence (and, for proposals, claims to feasibility), and match the author's own voice. It never changes a number, result, or citation, and it is not for evading AI-use disclosure. Use when editing AI-assisted academic prose or grant proposals.

Apply the bundled Chinese-English typography and design standard for copy audits, spacing, punctuation, or design QA; skip ordinary rewriting.

Independently audit an existing research package or manuscript for evidence integrity, reproducibility, and release readiness; the audit is read-only.

Build, update, audit, or query a source-backed project Wiki; skip ordinary documentation and one-off repository questions without a Wiki.

Design HTML-native visual artifacts with a Claude Design style workflow, including slide decks, interactive prototypes, landing pages, one-page documents, posters, HTML email and option canvases. Use when the user asks to design, mock up, prototype or present something visual. Skip small fixes inside an existing UI system.

Design a prospective experiment, benchmark, simulation, or evaluation protocol for a scoped research question; skip analysis of collected data.

Locate the cause of a non-obvious failure or evaluate code-review feedback against current evidence before changing behavior. Use for bugs, regressions, performance problems, flaky tests, or review comments that are ambiguous, conflicting, or technically uncertain; skip errors whose cause and safe correction are already explicit.

Build a reproducible corpus for systematic or scoping literature review, including bilingual search and screening; skip small claim checks.

Carry an accepted plan or known multi-step checklist through verified completion. Use when the direction is settled and the user wants implementation to continue without repeated approval; skip planning, diagnosis-only requests, and routine single-step answers.

Build new frontend apps, dashboards, games, creative websites, hero sections, and requested redesigns as working code. Grounds the design in the subject, plans tokens, builds code-native UI, and checks the rendered result once; Image Gen concepts are optional. Use for new builds or explicit redesigns; skip small fixes inside an existing design system.

Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.

Apply small, meaning-preserving rhetorical edits to a finished academic manuscript, to counter wording-driven LLM review penalties while keeping the scientific assessment a human could make materially unchanged. Uses model-agnostic strategies from LLM reviewer preference research without querying a target reviewer. Use after ordinary writing and polishing; not for drafting or generating reviews.

创作或深度改写自然、有依据、有中文韵律的文章与叙事。用于知乎长答、论坛帖、公众号、博客、评论、人物与行业稿、科普、故事、对白、口播和演讲，以及明确要求中文去 AI 味或活人感的改稿；保留事实、原意和语气，不补造来源或亲历。普通校对、技术文档、结构化报告和聊天回复使用 refine-text。

Keep long-running or scope-ambiguous work on the requested critical path. Use for open-ended improvement loops, SOTA or publication goals, or decisions about hashes, freezes, audits, migrations, compatibility, hardening, broad tests, dependencies, abstractions, and infrastructure when optional process may delay delivery; skip routine direct changes.

Turn an unclear goal or multi-part delivery into a lean executable direction. Use when objectives, constraints, trade-offs, or success criteria are unclear, or when delivery spans dependent steps or meaningful risk; skip routine direct changes with an obvious path.

Route academic and professional text polishing to academic-humanizer. Edit other supplied text within the user's requested scope while preserving meaning and citations; keep substantive reasoning with the current task.

Reverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.

Maintain one recoverable goal record when work must continue across sessions or be handed to another worker. Skip ordinary tasks unless the user explicitly asks for this record.

Start or resume a durable research program only when the user asks for one or the project already has .research-program state. Skip ordinary questions, one-off papers, and independent release audits.

用户明确说出 /team、$team 或“混动”时，按项目与变更范围进行隔离审查、协调裁决，并默认用 GPT-6 Luna Low 实施和监看。普通单模型任务不触发。

Audit and simplify an existing hierarchy of AGENTS.md, AGENTS.override.md, CLAUDE.md, or similar agent-instruction files. Use when removing duplicate, stale, vague, conflicting, or overly procedural rules; preserving directory-specific constraints; or deciding whether an instruction should be kept, shortened, rewritten, or deleted. Skip greenfield instruction design and unrelated routine edits.

规划、生成、修改和检查数学建模与科研图表；包含143个完整源码配方、3套组合模板、统一配色和来源差异工具，以及Draw.io/TikZ、HTML/Mermaid和科学场景插图。

Shape a response for low-friction reading and execution. Use only when explicitly invoked or when the user explicitly requests action-first, ADHD-friendly, low-cognitive-load, skimmable, step-by-step, or no-filler output. Preserve required detail, safety, agent autonomy, and the requested format.

Draft or substantively revise a complete research manuscript from traceable study evidence, with source-grounded scientific writing checks. Use for full-paper writing; use a narrower editing workflow for sentence-only polishing.

使用知乎开放平台搜索知乎和全网内容、获取热榜、调用知乎直答，或读取当前用户自己的知乎创作、关注与收藏。用户提到知乎搜索、社区观点、真实经验、热点、热榜、知乎直答、我的知乎内容、我的关注、我的收藏、开放平台、API、MCP、Access Secret，或要求查看、安装和配置知乎 Skill 时使用。深度研究优先返回搜索原始来源；本人数据只读取完成任务所需的最小范围。

Read the full body of one user-specified public Zhihu answer through zhihu-cli, with HTTP/completeness checks and auth stop rules; not search or crawling.