dotfiles-claude

Send peer-to-peer message and control to Claude Code sessions. Use when coordinating multiple simultaneous Claude sessions, or user says "dm another claude", "list claude sessions", "peek peer claude", "spawn claude in tmux", "monitor remote claude", "notify peer in tmux", "send to peer", "handover to claude", "inspect peer transcript", "orchestrate claude sessions".

Lark/Feishu (飞书) CLI tool. Use when the user asks to do anything on Lark/Feishu — send or read chat messages (IM), read or edit docs/sheets/slides/wiki, work with multi-dimensional tables (Base/bitable), manage calendar/meetings/video-conference, mail, contacts, tasks, approvals, OKR, drive files, minutes, whiteboards, or attendance. Triggers: "lark", "feishu", "send a feishu message", "peek lark group messages", "conclude last lark meeting".

Extract clean, complete markdown from any web page — articles, docs, READMEs, blog/social posts, academic papers. Also use as a fallback when curl returns noisy HTML or WebFetch returns truncated, summarized, or refused results.

Claude has no native audio or video understanding — this skill delegates that to Google Gemini 3.1 Flash Lite (cheap multimodal) on the ofox gateway. Use when the user asks to analyze / transcribe / describe / understand a sound, song, voice memo, or video, or asks what is said, sung, shown, or played in a media file. Triggers: "analyze this audio", "transcribe the song/speech", "what does this voice say", "describe/understand this video/mp4/clip", "what's the music/story in this MV", "listen to", "watch this".

Run long-running tasks under babysit — supervised background runner with cgroup-enforced memory/CPU caps and observability-stall kill. Use BEFORE any long-running task (>2 min), compute-intensive jobs (e.g. builds/benchmarks/renders/training), background work, or when the user says "run in background", "use babysit". This skill defines guardrails and the mandatory workflow, not just a how-to.

Audit agent-facing documentation (CLAUDE.md, SKILL.md, AGENTS.md, prompt markdown, agent reference doc, memory pages) for over-description and propose trims. Use proactively after writing or editing any agent-facing doc, or when the user says "trim this doc", "audit doc bloat", "this is over-explained".

Resource-aware pre-launch checklist for long-running or heavy tasks — prevents OOM, wasted compute, and daytime disruption. TRIGGER before: any `babysit run` of compute work, builds/benchmarks/renders/training jobs, large-scale scrapes or scripted browser runs, spawning 2+ parallel subagents, full-repo test runs, or any Bash call with `run_in_background=true` whose runtime is uncertain.

Write, refine, and validate a Chinese technical blog post for Zhihu (知乎). This skill should be used before drafting a long-form Chinese technical article — or when the user says "写博客", "知乎文章", "zhihu post", "write for zhihu".

High-quality C++ OOP coding style (archibate / parallel101 lineage) that overrides sloppy AI-default C++. Use this skill WHENEVER writing, editing, refactoring, or reviewing C++ code (.cpp / .h / .hpp / .cc / .cxx), designing C++ classes, interfaces, APIs, or libraries, or when the user mentions C++ design, OOP, design patterns, dependency injection, RAII, or "clean / modern C++". Apply it even when the user does not explicitly ask for a style: the default way models write C++ leans on free functions, public mutable state, raw new/delete, sentinel return codes, and long loose parameter lists — this skill replaces all of that with abstract-class-or-data-class design, dependency injection, type-rich APIs, value-based error handling, and RAII ownership.

Append a single bullet to staging memory. Use when a durable fact, knowledge, lesson, or pitfall (mistake-correction pattern) emerged mid-conversation that's potentially worth persisting into long-term memory. Also use when user corrected your mistake, says "remember X", "remember not to X", or "next time, do Y instead of X", "do not mistake on X again".

Programmatic/headless Claude Code usage — CLI flags, stream-json protocol, Agent SDKs, session management. Use before spawning Claude as a subprocess, testing skill or agent behavior, running "claude -p", or building with the Agent SDK.

Use MarkItDown to convert various files to Markdown. Use when converting PDF, Word, PowerPoint, Excel, images, audio, HTML, CSV, JSON, XML, ZIP, YouTube URLs, EPubs, Jupyter notebooks, RSS feeds, or Wikipedia pages to Markdown format. Also use for document processing pipelines, LLM preprocessing, or text extraction tasks.

Get a cross-model second opinion from gpt-5.5 on your current approach. Use before substantive work (writing, editing, committing to an interpretation), when stuck (errors recurring, approach not converging), when changing approach, or before declaring a task done. The reviewer sees your opening task plus recent work from the session transcript.

Inspect or debug web pages in a headless browser — fill forms, click buttons, take screenshots, test web apps, review frontend UI/UX aesthetics. Use for headless browser automation. This skill should be used when user says "test web app", "headless browser", "browser automation", "chromium --headless", "screenshot of this page", "look the page".

Add a note only visible to user. A private scratchpad not visible in agent context.

Form a minimal self-contained prompt to reproduce this session

One sentence output style

Legacy opt-in pueue workflow. Use ONLY when the user explicitly types "/pueue" or asks for pueue by name. For all other long-running / background tasks, use /babysit.

Connect to user's local Chrome session. Inspect and interact with web pages in a headful browser shared with user cookies and login states — navigate pages, click buttons, take screenshots, test web apps in headful Chrome the user could see in real-time. Use for headful browser automation. This skill should be used when user says "use my browser", "open this page", "in my Chrome", "fill the form", "test in real browser", "show in browser", "my login account". Request explicit user approval before use.

Spawn and interact with interactive CLIs (python, gdb, etc.) in tmux — send keystrokes, scrape pane output. This skill should be used when running any interactive CLI command, starting program requires interactive stdin / PTY / TUI render, or when the user says "split pane", "run in tmux", "start REPL", "spawn in PTY", "test TUI". Use instead of Bash for interactive programs.

Guide for writing ast-grep rules to perform structural code search and analysis. TRIGGER when a text-only grep would need 2+ alternations, regex capture groups, multi-line patterns, or filtering by syntactic role (call vs. import vs. definition). Also TRIGGER when renaming a symbol across call sites + definitions, finding all callers of a method on a specific type, matching code shapes, or writing codemods.

Measure code quality quantitatively — cyclomatic complexity, cognitive complexity, maintainability index. Use when evaluating whether code meets quality thresholds.

Query AI-generated overview of GitHub repositories via DeepWiki MCP. Use when needing a fast high-level understanding of a repository's architecture and design decisions without fetching full repository.

Prompt cache cost optimization. Generates periodic conversation turns to keep the 5-minute prompt cache warm during idle gaps. TRIGGER immediately after any Bash or Agent call with `run_in_background=true` on the official Anthropic runtime, or when the user says "cache keep-alive".

Web search with time-window, region/language, and `site:` filters, academic papers (arXiv/SSRN), PDF table/figure extraction, BibTeX, image search, web page reading, embeddings, reranking, classification, and deduplication (text or images) via Jina AI. Use when searching web content, finding academic papers, or extracting figures from PDFs. Prefer this over WebSearch for better results.

Build distinctive frontend interfaces that avoid generic AI-slop aesthetics. Use when the user asks to create or improve a web page or component, or says "create web page", "design UI/UX", "fix AI slop design".

Spawn 2-4 parallel forks answering the same question, then judge and pick the best. User-invoked via /forkpick only.

A long-form article / blog post — masthead, hero image placeholder, article body with figures and pull quotes, author byline, related posts. Use when the brief asks for "blog", "article", "post", "essay", or "case study".

A two-spread digital e-guide preview — page 1 is a cover (display title, author, "What's inside" stats, table of contents teaser); page 2 is a spread (lesson body with pull-quote and a step list). Lifestyle / creator brand tone. Use when the brief asks for an "e-guide", "digital guide", "lookbook", "lead magnet", "creator guide", "playbook", "PDF guide", or "电子指南".

Quarterly / monthly financial report — masthead with KPIs, revenue and burn charts, P&L summary table, top-line highlights, and an outlook paragraph. Use when the brief mentions "financial report", "Q3 report", "MRR review", "P&L", or "财报".

Create, repair, validate, preview, and package Codex-compatible animated pet spritesheets from character art, screenshots, generated images, or visual references. Use when a user wants to hatch a Codex pet, create a custom animated pet, or build a built-in pet asset with an 8x9 atlas, transparent unused cells, row-by-row animation prompts, QA contact sheets, preview videos, and pet.json packaging. This skill composes the installed $imagegen system skill for visual generation and uses bundled scripts for deterministic spritesheet assembly.

8-Bit Orbit — Pixel-art neon arcade aesthetic on a deep navy void. Anything that should feel like a CRT screen at 2am: cyberpunk, gaming, web3, indie dev tools, hackathon demos.

Biennale Yellow — Solar yellow on warm parchment with deep indigo serif and atmospheric sun-glow gradients. Anything that should feel like an art-biennale poster or a museum's annual programme: exhibition decks, arts-institution announcements, design conference brochures, curatorial pitches, literary publications, studio retrospectives.

BlockFrame — Neobrutalist deck with pastel-neon color blocks and chunky black borders. Anything that should feel pop-graphic and design-led: indie SaaS launches, agency credentials, creative reviews, brand redesigns.

Blue Professional — Cream paper background with electric cobalt blue accents; clean modern professional. Anything that should feel modern-considered and lightly authoritative: B2B SaaS pitches, consulting deliverables, advisory updates, investor reports.

Bold Poster — Editorial poster aesthetic with massive Shrikhand display and a single fire-engine red accent. Anything that should land like a magazine cover: brand manifestos, founder vision decks, editorial / cultural pitches, creative reviews.

Broadside — Dark editorial canvas with a single fire orange accent and bilingual Latin/Chinese type stack. Anything that should land like a broadside newspaper headline: brand manifestos, magazine and cultural pitches, design talks, bilingual EN/CN decks, founder vision statements.

Capsule — Modular pill-shaped cards on warm bone with a full pastel-pop palette. Anything that should feel modular, modern, and a little Y2K: lifestyle brands, creator portfolios, DTC launches, beauty / wellness, agency credentials.

Cartesian — Quiet warm-neutral palette with classical Playfair serifs; tasteful and unhurried. Anything that should feel quiet, considered, and grown-up: investment theses, white papers, advisory work, longform research, gallery / cultural decks.

Cobalt Grid — Electric cobalt italic serifs on a graph-paper canvas, anchored by stair-stepped pixel-glitch decorations and slim hairline rules. Anything that should feel like a quietly serious design / research bulletin, art publication, or curated trend report.