Agent Workflow Claude Skills
4 122 agent workflow skills from 665 authors. They configure the agents themselves: memory, prompts, context and other skills. Half of them fit into 1 844 tokens or less — that is what one costs your context window when the agent loads it. 778 ship runnable scripts rather than instructions alone. 5 of them cannot work without an MCP server, most often task. We also found 589 copies of these same skills sitting in other people's repositories — counted once here, not 589 times.
4 122 unique 665 authors 4 updated this month 506 from vendors
The checkpoint coordinator for /feature. Routed to by /feature once a writing-plans plan exists. Groups tasks into batches, delegates each batch to subagent-driven-development (fresh author agent per task, full review chain, fresh verification), then stops for a human checkpoint before the next batch. The checkpointed counterpart to /sprint's autonomous run.
The parallel fan-out primitive. Routed to by any skill or command that splits work across independent units and dispatches an agent per unit — subagent-driven-development, /sprint, parallel /review. It owns the dispatch/collect/funnel discipline: bound concurrency, isolate units, collect every result, dedupe overlap, and funnel through finding-triage then checkpoint-aggregator. Raw agent output is never consumed before the funnel runs; an agent that errors drops its unit without corrupting the batch.
The authoring gate for new skills. Routed to when the user invokes /new-skill "<gap>". Five gated phases — gap evidence, scope, authoring, self-review against the v2 house style, routing integration. A new skill is not written until an existing one is proven not to cover the gap, and not shipped until it carries gated phases, hard rules, and a routing entry. Every authored skill matches the v2 format (frontmatter name+description, # name, Pre-flight, Phase N · gate, Hard rules).
The implementation engine. Routed to by /sprint (full plan, autonomous) and by executing-plans (scoped batch, checkpoint-gated). One fresh subagent per task — test-first via tdd — followed by spec-compliance review, quality review, and fresh-run verification. No single context accumulates drift, and nothing is accepted on a subagent's word.
OPTIONAL per-task isolation for autonomous parallel work. Routed to only on explicit opt-in by subagent-driven-development or dispatching-parallel-agents, so parallel units mutate files without colliding. Stands up one worktree per unit, works it in isolation, then integrates each unit back onto the caller's working branch for the caller's single commit-gate + finishing-a-development-branch exit. Never on the default path; it does not bypass a gate or finish per unit.
Author a new codeArbiter skill: prove the gap is real, get the spec approved, then write it.
The parallel fan-out primitive. Routed to by any skill or command that splits work across independent units and dispatches an agent per unit — subagent-driven-development, /sprint, parallel /review. It owns the dispatch/collect/funnel discipline: bound concurrency, isolate units, collect every result, dedupe overlap, and funnel through finding-triage then checkpoint-aggregator. Raw agent output is never consumed before the funnel runs; an agent that errors drops its unit without corrupting the batch.
The authoring gate for new skills. Routed to when the user invokes /new-skill "<gap>". Five gated phases — gap evidence, scope, authoring, self-review against the v2 house style, routing integration. A new skill is not written until an existing one is proven not to cover the gap, and not shipped until it carries gated phases, hard rules, and a routing entry. Every authored skill matches the v2 format (frontmatter name+description, # name, Pre-flight, Phase N · gate, Hard rules).
The implementation engine. Routed to by /sprint (full plan, autonomous) and by executing-plans (scoped batch, checkpoint-gated). One fresh subagent per task — test-first via tdd — followed by spec-compliance review, quality review, and fresh-run verification. No single context accumulates drift, and nothing is accepted on a subagent's word.
Author a new codeArbiter skill: prove the gap is real, get the spec approved, then write it.
Trim transcript clutter to extend session lifetime — analyze, prune a copy, or toggle the after-each-turn service. Dry-run by default; gains land at resume/compaction, not the current turn.
The checkpoint coordinator for /feature. Routed to by /feature once a writing-plans plan exists. Groups tasks into batches, delegates each batch to subagent-driven-development (fresh author agent per task, full review chain, fresh verification), then stops for a human checkpoint before the next batch. The checkpointed counterpart to /sprint's autonomous run.
The implementation engine. Routed to by /sprint (full plan, autonomous) and by executing-plans (scoped batch, checkpoint-gated). One fresh subagent per task — test-first via tdd — followed by spec-compliance review, quality review, and fresh-run verification. No single context accumulates drift, and nothing is accepted on a subagent's word.
The authoring gate for new skills. Routed to when the user invokes /new-skill "<gap>". Five gated phases — gap evidence, scope, authoring, self-review against the v2 house style, routing integration. A new skill is not written until an existing one is proven not to cover the gap, and not shipped until it carries gated phases, hard rules, and a routing entry. Every authored skill matches the v2 format (frontmatter name+description, # name, Pre-flight, Phase N · gate, Hard rules).
OPTIONAL per-task isolation for autonomous parallel work. Routed to only on explicit opt-in by subagent-driven-development or dispatching-parallel-agents, so parallel units mutate files without colliding. Stands up one worktree per unit, works it in isolation, then integrates each unit back onto the caller's working branch for the caller's single commit-gate + finishing-a-development-branch exit. Never on the default path; it does not bypass a gate or finish per unit.
Decide which model (Haiku/Sonnet/Opus) and effort level each subagent gets, when to cascade cheap-first behind a verifier, and how to run improvement loops safely (evaluator-as-selector, stop on regression). Covers the Managed Agents effort mechanics (per-agent effort levels, cost lever) and how to watch a subagent fan-out stream live so loop and escalation gates have something to observe. Use when spawning subagents via the Agent or Workflow tools, when fanning out more than a handful of agents, or when the user asks which model or effort a task should route to. Routing heuristics grounded in measured calibration data (references/calibration-2026-07-15.md), not vibes; the Managed Agents API specifics are operational, not calibrated.
Control Spotify playback and manage playlists via MCP server. Use when user requests playing music, controlling Spotify, creating playlists, searching songs, or managing their Spotify library.
Universal environment variable loader for AI agent environments. Loads secrets and config from Claude.ai, Claude Code, OpenAI Codex, Jules, and standard .env files.
Build and cache a personalized container environment from a Dockerfile-like spec. Supports both single-layer (one Containerfile -> one cached tarball) and multi-layer composition (compose [base, scientific, mojo, ...] into one container with each layer cached independently). Use when the user mentions "container layer", "Containerfile", "custom container", "environment setup", "cache my installs", "uv shim", "composable layers", or wants to persist package installations, skills, or environment config across ephemeral sessions. Also triggers when the user asks to snapshot, restore, or rebuild their environment, or wants to capture ad-hoc package installs into a reproducible spec.
Generate optimized instructions for Claude (Project instructions, Skills, or standalone prompts). Use when users request creating project setups, writing effective prompts, building Skills, or need guidance on instruction types for Claude.ai.
Creates production-ready MCP servers using FastMCP v2. Use when building MCP servers, optimizing tool descriptions for context efficiency, implementing progressive disclosure for multiple capabilities, or packaging servers for distribution.
Builds a portable, embedding-free knowledgebase from a set of files and delivers it as a self-contained `.skill` bundle (BM25 index + bundled searcher + query protocol). Use when a user wants to turn uploaded files, a folder, or a corpus into a searchable knowledgebase they can hand to any agent — phrased as "make a knowledgebase", "build a KB skill", "package these docs for retrieval", "create a searchable bundle", or references to a `.skill` KB. The output runs anywhere with Node or Python — no model, no install, no network. Distinct from `bm25` (ephemeral in-session search) and `building-github-index` (markdown project-knowledge index).
Creates Skills for Claude. Use when users request creating/updating skills, need skill structure guidance, or mention extending Claude's capabilities through custom skills.
>- Distill Opus-level reasoning into optimized instructions for Haiku 4.5 (and Sonnet). Generates explicit, procedural prompts with n-shot examples that maximize smaller model performance on a given task. Use when user says "down-skill", "distill for Haiku", "optimize for Haiku", "make this work on Haiku", "generate Haiku instructions", or needs to delegate a task to a smaller model with high reliability.
Discover and load skills on demand from /mnt/skills/user/. Use when you need a capability but don't know which skill provides it, when the boot-emitted skill list is names-only and you need a full description, or when you want to list the catalog. Verbs are list (names only), search (rank by name/description match against a query), and show (emit the full SKILL.md for a named skill).
Break out of a locked problem frame by picking one disciplined move — reframe, provocation (Po), random stimulus, SCAMPER, inversion, perspective shift, constraint play, or family traversal — and committing to it before evaluating. Use when stuck, when options feel narrow or obvious, when iterations produce variations of the same idea, or when the user says "widen this", "break out of", "think differently", "I'm stuck", "feels too obvious", "stress-test the framing", "what am I missing", or holds two related examples and asks what lies between or beyond them. Complements challenging (which evaluates) and convening-experts (which synthesizes viewpoints); this skill generates distance, not judgment.
Install skills from github.com/oaustegard/claude-skills into /mnt/skills/user. Use when user mentions "install skills", "load skills", "add skills", "update skills", "refresh skills", or references a skill not currently installed.
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.
Disciplined, validation-gated revision of an EXISTING skill so each edit is a measured improvement rather than a guess. Use when editing, revising, or tuning a skill that already exists and there is evidence it underperforms (observed failures, drift, complaints) — invoke by name, or have versioning-skills / creating-skill defer to it before applying edits. Not for authoring a brand-new skill from scratch (use creating-skill) or one-off prose.
Orchestrates parallel API instances, delegated sub-tasks, and multi-agent workflows with streaming and tool-enabled delegation patterns. Routes by surface — native subagents in Cowork and Claude Code, httpx fan-out on claude.ai — and covers Gemini delegation via the Cloudflare AI Gateway on every surface. Use for parallel analysis, multi-perspective reviews, or complex task decomposition.
>- Skill-aware orchestration with context routing. Decomposes complex tasks into skill-typed subtasks, extracts targeted context subsets, executes subagents in parallel, and synthesizes results. Self-answers trivial lookups inline. No SDK dependency — uses raw HTTP via httpx. Use when tasks require multiple analytical perspectives, when context is large and subtasks only need portions, or when orchestrating-agents spawns too many redundant subagents.
Exhaustive problem space exploration using the MIT Synthetic Neurobiology "tiling tree" method. Partitions a problem into MECE (Mutually Exclusive, Collectively Exhaustive) subsets recursively via parallel subagents, then evaluates leaf ideas against specified criteria. Use when users say "tiling tree", "tile the solution space", "exhaustively explore approaches to", "what are all the ways to", or request a MECE breakdown of a problem. Requires orchestrating-agents skill.
File-upload bridge for Claude Code on the Web. CCotw has no native file mount; this skill creates a throwaway GitHub branch the user can drop files onto via the github.com web UI, then fetches them locally on the next turn. Use when the user wants to upload, share, or send files into the session, or when a task clearly needs files the user has on disk that aren't in the repo.
REQUIRED for all skill development. Automatically version control every skill file modification for rollback/comparison. Use after init_skill.sh, after every str_replace/create_file, and before packaging.
DEPRECATED - Use browsing-bluesky skill instead. Sample and analyze Bluesky firehose to identify trending topics and content clusters. Use when user asks about "what's happening on Bluesky", "Bluesky trends", "zeitgeist", "firehose analysis", or wants to see real-time topic clusters from the network.
Embedded hardware debugging workflow for probe-rs targets using embedded-debugger-mcp. Use when Codex or Claude Code needs to inspect debug probes, validate embedded debugger setup, start the MCP server, guide a user through ARM Cortex-M/RISC-V flashing/debugging/RTT workflows, or operate without installing an MCP client by using the CLI plus prompts.
Scan current session for skills that need creating or updating, then apply changes
Save session state and generate a return prompt (quick or full mode)
Turn a session into a high quality Claude skill with research and planning
Session retrospective - analyze what happened, find friction, auto-implement quick wins, publish report
Socratic teaching loop for any Claude Code session — quiz yourself on what actually happened, confirm mastery item by item, and don't finish until everything's locked in
> Orchestrates the discovery, mapping, design, and implementation of a premium Flutter-based frontend for an Agent Development Kit (ADK) agent written with Python.
>- Apply the SAIL (Secure AI Lifecycle) V2 framework by Pillar Security to secure AI applications and agents. Use this skill whenever the user asks about AI or agent security — assessing an AI system or agent architecture for risks, building an AI security roadmap or maturity assessment, writing or reviewing an AI security policy, creating compliance checklists (ISO/IEC 42001, EU AI Act, OWASP LLM/Agentic, DASF, AIUC-1), prioritizing AI security controls, evaluating AI vendors or tools (RFPs, security questionnaires), securing MCP servers, agent identities, or LLM apps, or asking what "SAIL" or a "SAIL ID" (e.g., SAIL 5.17) means. Also use it when the user is building agentic features and wants to know which security risks apply — even if they never say "SAIL".
異なるproviderのAI（Codex/Claude/Grok/Cursor/Antigravity等）から独立意見を集め、共通論点で相互検討して統合する。最初に利用可能エージェントを検出する。「みんなで考えて」「マルチエージェントで分析して」「複数のAIでレビューして」と言われたら使う。
スキルの作成・改善を行うスキル。既存スキルの分析から抽出したパターンとテンプレートで統一感のあるスキルを効率的に作成。スキルを作りたい、新しいスキルを追加したい、スキルを改善したい、SKILL.mdを書きたい場合に使用。
Author and validate Vally evals for Agent Skills under .github/skills. WHEN: "write a skill eval", "add trigger eval", "test skill routing", "add anti-trigger tests", "create skill capability eval", "harden skill graders". DO NOT USE FOR: single MCP tool prompt-to-tool coverage (use eval-authoring-tool), multi-tool or multi-turn scenarios (use eval-authoring-workflow), authoring SKILL.md itself (use skill-authoring).
Analyze markdown files for token efficiency and reduce context-window bloat. **UTILITY SKILL**. DO NOT USE FOR: code optimization, general file editing, non-markdown files. TRIGGERS: optimize markdown, reduce tokens, token count, token bloat, too many tokens, make concise, shrink file, file too large, optimize for AI, token efficiency, verbose markdown, reduce file size. INVOKES: waza CLI.
Author and validate hermetic single-tool Vally evals under evals/tools. WHEN: "write a tool eval", "add prompt-to-tool coverage", "test MCP tool selection", "add tool catalog eval", "create single-tool scenario", "harden tool-call grader". DO NOT USE FOR: skill routing or capability evals (use eval-authoring-skill), multi-tool, multi-turn, or live scenarios (use eval-authoring-workflow).