查找技能

Runs a long swarm of hundreds to thousands of cheap headless coding workers that drive a large failing-test backlog to green. Failures are clustered by signature, each worker fixes one cluster in its own git worktree, a serial no-regression merge gate lands the fixes, the loop resumes across restarts for hours, and it federates across sibling sessions over git branches when one container's CPU is not enough. Use for "fix these thousands of failing tests with a Haiku swarm", "drive the test suite to green with hundreds of agents", "build this compiler or interpreter against a conformance suite", "launch, relaunch or check in on the swarm loop", "the swarm is stuck, idle or rejecting everything", "parallel agents fixing tests keep hitting merge conflicts and regressions", "add sibling containers to the swarm", or "compare models on the tail the swarm could not fix". Ships the harness, a per-project TOML config, a fake-worker selftest, check-in signals and the 2026-10-12 haiku-cc failure catalogue.

中文简介正在准备中。当前显示原文说明。

Makes small documents as shareable Tiny Little Office links, where the URL itself is the document. Covers slide decks, posters, zines, spreadsheets, charts and small-business models, letters, résumés, invoices, outlines, diagrams (mind map, org chart, family tree, swimlanes, timeline, Venn, seating, network), drawings, pixel art, animations, calendars, Gantt schedules, trip maps and a calculator that shows its working. Use when someone wants such a document as a link they can open, edit and pass on rather than a file, or names tinylittleoffice / TLO.

中文简介正在准备中。当前显示原文说明。

Builds a product description repo for a software product - prose documents that say, feature by feature and event by event, what the user sees, what they can do, and exactly what happens when they do it, including when they stop halfway. Drafted from the code and tests, verified against the running product with hand checklists, and triaged into a bug list. Works for canvas editors, web apps, CLIs, chat products, mobile apps, and libraries or SDKs. Use when asked to "write a product description for X", "describe the user experience of X", "document how X behaves for the user", "make a behavior-spec repo", or for an event-by-event account of an app's behavior rather than API docs. Also resumes or extends an existing product description repo.

中文简介正在准备中。当前显示原文说明。

Fans Claude work out over the Anthropic API from surfaces with no native subagents, chiefly claude.ai chat and projects, with parallel calls with a shared cached context, streaming, long-lived conversation threads with cached history, and named agent pools; also delegates to Gemini through the Cloudflare AI Gateway on any surface. Use for multi-perspective analysis, a panel of reviewers, or task decomposition from claude.ai, or for Python code that calls several Claude instances concurrently over the API.

中文简介正在准备中。当前显示原文说明。

Chooses how to launch subagents from Claude Code (the Agent tool, a Workflow script, or headless claude -p children) and runs a headless fan-out from one background command. Use for "fan this out to 50 agents", "run these tasks in parallel", "launch subagents for each file", "should I spawn agents or write a workflow", "batch this through Haiku", "run claude -p in parallel", "dispatch a task list without the launches eating my context", or before any launch of more than a handful of subagents, where the launching itself is the cost. Covers the parent-turn cost of each mechanism, concurrency ceilings, what context and tools each child gets, effort control, output contracts, a cost ledger, and resume after interruption.

中文简介正在准备中。当前显示原文说明。

Routes, triages, flags and rates a piece of text with a probability for every option: which department or queue a ticket goes to, which intent a message expresses, whether a yes/no condition holds (is this tool call grounded, should the agent ask first, is the customer angry), and how severe or urgent it is on a rubric, each with a calibrated confidence. Runs Claude Haiku in the request and response shapes of OpenAI's Decisions API, for when no purpose-built decision model is at hand. Use for "route this ticket", "triage these", "classify with a confidence", "give me a probability that", "rate the severity", "score against this rubric", guardrail checks before a tool call, or "a Decisions API without OpenAI".

中文简介正在准备中。当前显示原文说明。

Query, filter, and transform Markdown structurally with mq — a jq-like CLI for Markdown. Use to extract headings/sections/code-blocks/links from .md files, build a table of contents, pull code blocks of a given language, slice or reshape LLM prompt/output Markdown, or batch-transform docs. Triggers on "extract sections from this markdown", "get all the code blocks", "jq for markdown", "mq", or any structural query over Markdown that grep/Read can't do cleanly.

中文简介正在准备中。当前显示原文说明。

Adds a searchable text layer to a scanned PDF with ocrmypdf. Installs the toolchain at runtime in ~18s. Use when a PDF's pages are images and the deliverable is a file to keep, grep, or hand to another tool: 'make this PDF searchable', 'OCR this scan', 'I can't select the text in this PDF', 'search across these scanned pages', 'the pdf skill returned nothing'. Handles non-English scans via tesseract language packs. NOT for charts, diagrams, handwriting, or slide layouts. Route those to transcribing-images.

中文简介正在准备中。当前显示原文说明。

Apply semi-formal certificate reasoning to code analysis — patch verification, fault localization, patch equivalence. Use when reviewing patches, hunting bugs across scopes, comparing fixes, or when code reasoning requires tracing execution across files/modules. Triggers on code review, bug localization, patch comparison, name shadowing, scope analysis, regression checking.

中文简介正在准备中。当前显示原文说明。

Execute programs on a compiled transformer stack machine where every instruction fetch and memory read is a parabolic attention head. Demonstrates that transformer attention + FF layers can implement a working computer. Use when user mentions "llm-as-computer", "lac", "stack machine", "compiled transformer", "percepta", "parabolic attention", "execute program", or asks to run/trace programs on the transformer executor.

中文简介正在准备中。当前显示原文说明。

Generate navigable semantic maps from PDF documents. Extracts section structure via font analysis, then runs LLM extraction per section for claims, symbols, and dependencies — all page-anchored. Produces _MAP.md (progressive disclosure), .symbols.json (definition index), .anchors.json (claim references), and a _USAGE.md snippet for CLAUDE.md. Use when analyzing papers, specs, or legal docs; when asked to "map this document", "index this PDF", "what does this paper say"; or when a coding agent needs grounded reference material from a PDF source. Analogous to tree-sitting but for prose documents.

中文简介正在准备中。当前显示原文说明。

DEPRECATED, renamed to orchestrating-chat-agents. Kept so existing imports of its scripts keep working; every module here loads the same module from orchestrating-chat-agents. Use orchestrating-chat-agents for parallel Claude calls over the API from claude.ai, and dispatching-subagents for fan-out in Claude Code.

中文简介正在准备中。当前显示原文说明。

Interactive codebase orientation for a HUMAN who wants to learn the code. Runs the same tree-sitting + featuring pipeline as exploring-codebases but synthesizes it into guided exercises and an HTML teaching artifact rather than an analysis document. Use for "orient me to this repo", "teach me this codebase", "walk me through this code", "learning orientation", or when someone wants genuine comprehension rather than a task completed. The audience is the test: if nobody is being taught and the goal is to get work done, use exploring-codebases instead.

中文简介正在准备中。当前显示原文说明。

Semantic Python code queries via a stdio LSP client driving pyright-langserver. Provides binding-resolved go-to-definition, find-references, hover types, type diagnostics, file symbol outlines, and project-wide symbol search — name resolution and type inference that tree-sitter and ripgrep cannot do. Use when you need to follow an import to a definition, find all real uses of a symbol (excluding same-named-but-unrelated ones), get an inferred type, surface type errors, outline a file, or search symbols across a project. Triggers on "go to definition", "find references", "what type is", "resolve this symbol", "symbol outline", "find symbol in project", "pyright", "type-check this file".

中文简介正在准备中。当前显示原文说明。

Enables GitHub repository operations (read/write/commit/PR) for Claude.ai chat environments. Use when users request GitHub commits, repository updates, DEVLOG persistence, or cross-session state management via GitHub branches. Not needed in Claude Code (has native git access).

中文简介正在准备中。当前显示原文说明。

Generate guardrailed UI from natural language. Emits constrained JSON that a Preact runtime renders. Use when the request is for a dashboard with metrics, charts, or tables; an admin panel; a data visualization interface; or a form-based application.

中文简介正在准备中。当前显示原文说明。

Discovers and indexes Python code in skills, enabling cross-skill imports. Use when importing functions from other skills or analyzing skill codebases.

中文简介正在准备中。当前显示原文说明。

Use when the user explicitly asks to use the informed-patient skill to prepare for a medical appointment, organize symptoms before seeing a doctor, or evaluate the evidence behind a diagnosis or treatment. Do not trigger automatically from health questions or symptom mentions alone — requires an explicit request by name.

中文简介正在准备中。当前显示原文说明。

Install and drive Google's Antigravity CLI (`agy`) as a non-interactive sub-agent. Use when orchestrating agy, running Antigravity agents from a script or sandbox, delegating a task to Google's agent harness, or wanting a Gemini-backed peer agent alongside Claude.

中文简介正在准备中。当前显示原文说明。

Multi-conversation methodology for iterative stateful work with context accumulation. Use when users request work that spans multiple sessions (research, debugging, refactoring, feature development), need to build on past progress, explicitly mention iterative work, work logs, project knowledge, or cross-conversation learning.

中文简介正在准备中。当前显示原文说明。

Listen to generated music by measuring it and by reading it as sheet music: render Strudel code or record a Web Audio page through the real engine in headless Chromium, then read spectrograms, pYIN pitch lines, chroma, roughness and semitone-rub meters; engrave notes as a score with beat-by-beat harmony; transcribe a melody from audio; read MusicXML/MIDI/ABC or a score image into notes and into Strudel code. Use to check whether generated music sounds right or better (before/after a change), to match a reference track, spectrogram image or score, or to find discordant chords, a buried melody, clipping or muddy mixes. Triggers on 'does it sound better', 'listen to', 'render and check', 'spectrogram', 'sfft/stft', 'discordant', 'clashing chords', 'sounds off', 'compare to the original', 'reproduce this track in Strudel', 'sheet music', 'score', 'notation', 'MusicXML', 'MIDI', 'ABC', 'transcribe', 'OMR'. Pairs with strudeling (writing Strudel); for format conversion use processing-video.

中文简介正在准备中。当前显示原文说明。

Delivers a static Hello World HTML demo page with bookmarklet. Use when user requests the hello demo, hello world demo, or demo page.

中文简介正在准备中。当前显示原文说明。

Invokes Google Gemini models for structured outputs, image generation, text-to-speech narration, multi-modal tasks, and Google-specific features. Use when users request Gemini, image generation, Gemini TTS or a synthesized voice, structured JSON output, Google API integration, or cost-effective parallel processing.

中文简介正在准备中。当前显示原文说明。

Generates WAFFLES Declarations for social media posts — preemptive lists of what a post does NOT say. Use when users mention WAFFLES, ask for clarifications on their post, want to prevent misinterpretation, or request disclaimers for controversial/nuanced takes.

中文简介正在准备中。当前显示原文说明。