skills

Read-only codebase health audit — MECE pillars architecture (A1–A6), code (C0–C3), security (S1), verification (V1–V2), dependencies (D1), operations (O1). Chat-only P0/P1/P2 roadmap; effort snapshot, standard, or deep; variants simplicity, anti-pattern, branch, direction, cicd, and pillar-scoped review. Use for codebase audit, architecture review, security hygiene, CI/CD review, over-engineering and anti-pattern scan; not article writing, topic learning, implementation edits, or plans/.

Guard long, ambiguous, or stateful AI-agent work from drift. Use for multi-step implementation, compaction at phase boundaries, task contracts, handoff/resume, or work spanning many tool calls, files, or sessions — not single-shot article writing, topic learning, or codebase audit.

Remove AI-writing tells from finished prose only — neutral tone, preserve facts and structure. Optional deep QA (tell-focused L1–L4). Use for 去AI味, humanize, de-AI polish, 润色; not 审校定稿 with author voice-default, publish-ready writing pipeline, blank-page drafts, or chapter reorder.

Fetch a random classical Chinese poem line from the Jinrishici HTTP API and format with standard Chinese attribution punctuation (author, title). Use for daily poem, briefing garnish, 今日诗词.

Log in via Keep Open API and fetch runs into Garmin-compatible running.json (VDOT, training load, period stats, optional per-km segments). Use for Keep run sync, fetch_keep_run, running.json, or exporting Keep running data.

Audit, design, categorize, distribute, and measure agent skills using lessons from Anthropic's Lessons from building Claude Code: How we use skills. Use when reviewing an existing skill, deciding whether a workflow deserves a skill, planning a skill library, turning team knowledge into skills, choosing skill categories, writing trigger descriptions, designing progressive disclosure, or planning skill marketplace and usage measurement.

Fetch URL or PDF as clean Markdown for reading, citation, or archive (读). Modes convert (default), summarize, save, preview. Routes CN platforms, GitHub, PDF, proxy cascade, MCP fallback. Use for 读链接, 抓取正文, 提取文章; not article drafts, topic learning workspace, or codebase audit.

Article and document writing lifecycle (写) — ideate, gather, transcribe, compose, refine with author voice. Long-form, tutorials, scripts, talks, Xiaohongshu. Default voice in voice-default.md. Use for 写文档, 写文章, 选题, 审校定稿, publish-ready; not neutral-only de-AI on finished text, URL ingestion, topic learning, or codebase audit.

Learn one topic end to end (学): scoped mission, layered sources, mastery map, verified evidence, canonical article, interview layer, learning records. Modes new or refresh. Use for 主题学习, 技术深挖, 溯源验证; not single-link summaries, blank-page article writing, or codebase audit.

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.