Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Chooses the right FORMAT for instructions on Claude.ai — project instructions, a skill, or a standalone prompt — and gives the format-specific structure once chosen. Use when the question is which container an instruction belongs in ("should this be a skill or project instructions", "where do I put this", "how do I set up this project", "is this worth a skill"), or when someone has instructions and does not know how to package them. For the writing principles that apply inside any of the three formats, use writing-instructions. For building, testing and packaging a complete skill directory, use creating-skill.
Penetration testing efficiency rules. Prevent inefficient behaviors such as blind enumeration and tool blocking, and ensure the maximum number of vulnerabilities is found within a limited time. Use when starting a penetration test, when a scanning tool is taking too long, when stuck in a blind brute-force loop, or when you need to decide the next attack direction. Recommended reading before each round of penetration testing
Starting point for exploring how a PostHog MCP server's tools are used — routes a broad question to the typed tool that answers it. Use when the user asks "how is my MCP doing?", "what should I look at?", "explore my tool calls", "who uses my MCP tools?", "what are agents doing with the MCP?", or pastes an MCP analytics URL without a specific question. Offers a menu of questions, each backed by a query tool, then hands off to the focused skill.
Generates production-ready FastGPT workflow JSON from natural language requirements. Uses AI-powered semantic template matching from built-in workflows (document translation, sales training, resume screening, financial news). Performs three-layer validation (format, connections, logic completeness). Supports incremental modifications to add/remove/modify nodes. Activates when user asks to "create FastGPT workflow", "generate workflow JSON", "design FastGPT application", or mentions workflow automation, multi-agent systems, or FastGPT templates.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Install, configure, and troubleshoot Citedy's game-sounds feedback audio for Claude Code and supported CLI environments. Use when a user wants coding-agent event sounds, sound-pack rotation, volume or event toggles, playback checks, custom packs, or the game-sounds CLI/plugin. Triggers on: game-sounds, coding sounds, Claude hook sounds, task-complete sound, sound pack rotation, or @citedy/game-sounds.
Use when asked to set up, install, or refine an agent quality framework in a repository (rules with gates, architecture contracts, proven-red tests, ratchets, playbooks), or when the user says "gap-trap", "gap-trap setup", "gap-trap refine", "gate the rules", or "make the agents' code match the rules".
Orchestrate software engineering workflows across the canonical get-fable coding lifecycle with deterministic routing and evidence precedence. Use when starting a complex coding task, navigating lifecycle phases, resuming work with durable .fable state, or routing between research, planning, testing, verification, and recovery—even if the user doesn't explicitly say "get-fable" (e.g. "follow fable lifecycle", "orchestrate this project", "what is the next engineering step", "route my task"). Do NOT use when an individual specialist skill already clearly owns a bounded subtask.
Add or audit GitHub Copilot **personal** custom instructions — the per-user prompt configured at github.com/copilot that applies to every Copilot Chat conversation that account has on github.com. Covers where to set them, what they do and don't apply to (github.com Chat only — not IDE, not mobile, not coding agent), how they merge with repository and organization instructions, and verification. Use when the user says "set personal Copilot instructions", "configure my Copilot", "Copilot keeps responding in the wrong language", "make Copilot use my preferred tone", "personal Copilot prompt", "user-level Copilot config", "Copilot is ignoring my preferences", or invokes /github-copilot-personal-instructions. Do NOT use for repository-scoped instructions — that is /github-copilot-repo-instructions.
Side-route orientation skill. Use when the user asks 'what's next', 'where am I', or is returning to a repo mid-pipeline and wants to know which skill to run next. Reads repo state (branch, PRs, issues, research archive, milestones) and recommends one next step with a one-line reason. Not for executing work — only for routing.
Turn a vague idea or feature request into a written, reviewable spec and implementation plan before any code is written. Use when the user says "I want to build X", "help me think through", "I have an idea", or hands over an underspecified request.
Analyze Claude Code session logs and surface productivity improvements. Extracts thinking blocks, tool usage stats, error patterns, debug trajectories - then generates friendly, actionable recommendations. Triggers on: introspect, session logs, trajectory, analyze sessions, what went wrong, tool usage, thinking blocks, session history, my reasoning, past sessions, what did I do, how can I improve.
INVOKE THIS SKILL when implementing human-in-the-loop patterns, pausing for approval, or handling errors in LangGraph. Covers interrupt(), Command(resume=...), approval/validation workflows, and the 4-tier error handling strategy.
Invoke this skill when your LangGraph needs to persist state, remember conversations, travel through history, or configure subgraph checkpointer scoping. Covers checkpointers, thread_id, time travel, Store, and subgraph persistence modes.
Always-on Light skill for runtime project memory and project management: land "how far the project got / what was decided / which version shipped / how terminology is unified / where we left off" in each project's own .light/ directory (project card + decision log + version history + controlled glossary + cross-session handoff card), reusing the passport.py engine to manage the DAG ledger (no reinvention). It owns the consistency source of truth: change a definition and the change broadcast rescans all materials; it is the memory substrate for the controller's "automatically report where we left off at session start". When to use: long-running/cross-session projects, when background/progress/version/decisions must be remembered, resuming or taking over at session start, logging important progress immediately, handing off as context runs out, after changing terminology/metric/innovation-point definitions. Trigger words: memory / project management / project card / decision log / version history / version record / resume / take over / where we left off / next step / handoff / startup prompt / glossary / change broadcast / archive / ledger / .light / passport / cross-session. Core discipline: memory goes into explicit .light/ files (not vector retrieval); convert relative dates to absolute; external mutable facts carry [snapshot]; for self-consistency use pm.py audit and its exit code (don't just say "it's recorded"); irreversible decisions (archiving / writing back lessons) stop and ask the user.
OpenClaw local AI assistant stack. Covers architecture, tools, gateway operations, channels, and onboarding. Use when deploying, configuring, or operating an OpenClaw instance, managing gateway routing, setting up channels, or working with the multi-agent tool governance system. Keywords: OpenClaw, gateway, tools, channels, agents.
Use Opper from application code and over HTTP, and as the source of truth for Opper platform concepts: the gateway, control plane, models, the OpenAI / Anthropic / Google / OpenResponses-compatible endpoints under /v3/compat, structured output, server-side tools (web search), roundtable, streaming, tracing, and building agents. Use whenever the user writes code against Opper in any language (the OpenAI, Anthropic or Google Gen AI SDK, or an agent framework, pointed at /v3/compat), asks about Opper concepts, models, endpoints or raw HTTP, or migrates from OpenAI / OpenRouter / Anthropic, off Opper's legacy /v3/call (opper.call), or off the discontinued `opperai` SDK, even if they only say "Opper". For media generation (images, audio, video, OCR), files, and realtime voice, use `opper-multimodal`. For endpoint signatures, fetch the live spec at https://api.opper.ai/v3/openapi.yaml
Dispatcher for the creative suite: diagnoses a multi-asset creative request with a few questions, then routes it through the right sequence of creative skills, naming the artifact each step passes to the next, and wires in the FAL.ai automation (single model fal-ai/nano-banana-pro via docs/creative_cli.py) when available. Use when the user asks for several creative deliverables at once or an outcome that implies them — 'assets for my product launch', 'set up my brand visuals', 'I need a logo, product shots, and social posts', 'a week of content', 'rebrand everything' — or when it's unclear which creative skill applies. Produces a Creative Production Plan: chosen route, ordered skill sequence with hand-off artifacts, generation checklist, and asset output map. For a single obvious asset, skip this and use the specific skill directly.
Configure local operator assumptions for working with Paperclip through paperclipai and Paperclip MCP fallback. Use when setting up this repo, checking Paperclip CLI context, installing project-local or explicitly global MCP config, or preparing the Paperclip operator skill suite.
Fresh-context review of staged changes, branches, PRs, or file sets. Delegates to a fresh reviewer; routes to external CLIs (Copilot, Codex) via --model. Use when: user says "peer review" ("peer review PR 5", "peer review staged"), "fresh review", "another set of eyes", "sanity check", "quick review before I push", "review with Copilot", or "review using Codex". Do NOT trigger on bare "review" phrases ("review my changes", "review PR N", "review staged") — those route to code-review.
Global master controller for the product manager Skill library. Manages 20 PM Skills: 13 execution tools (PRD, review, prioritization, roadmap, data analysis, experiments, event tracking, surveys, competitive analysis, retrospectives, prototyping ×3) + 7 expert advisory groups (Cagan, Torres, Yu Jun, Mom Test, Story Mapping, Build Trap and their master controller). Capabilities: (1) automatically route to the most suitable Skill based on the problem the user describes (2) orchestrate multi-Skill workflow chains, passing the previous step's output to the next (3) hand judgment-type questions to pm-advisory-board for expert review. Trigger words: "pm-master", "product master control", "which Skill should I use", "help me push this requirement forward", "walk through it from start to finish", "full chain", or when the user describes a product work scenario without specifying which Skill to use, or when a task clearly requires multiple Skills to be completed in sequence, use this Skill first.
Professor Skill creates a university-course skill from slides, syllabi, exams, transcripts, notes, and chat logs. Use when the user wants a review-first, exam-focused, teacher-style skill that models how a professor highlights topics, writes questions, and deducts points.
Create high-quality prompts or improve existing ones — meta-prompting with task decomposition, expert personas and a verification checklist. Use when the user says 'tạo prompt', 'viết prompt cho tôi', 'giúp tôi viết prompt', 'cải tiến prompt này', 'prompt này chưa tốt', 'review prompt', or needs a prompt for system/task/conversation/role. Basic version — use prompt-generator-v2 (KERNEL) if you need a stricter structure and prompt chaining.
Use when building ANYTHING - the universal reference-first system that routes every task to proven reference implementations instead of generating from assumptions. Covers API design, database schemas, testing strategies, CI/CD pipelines, code patterns, DevOps, and any other domain where professional reference implementations exist
Use when the user invokes /reflect or asks to reflect on, wrap up, or take stock of the current session. Re-reads the in-context conversation for anything that would die when the session closes—dropped action items, durable lessons with no home, stale docs, skillify opportunities, global/project CLAUDE.md and AGENTS.md and README gaps, and follow-ups worth an issue or a PR—surfaces them as one ranked shortlist, then acts on the ones the user selects. Not a git-history retro and not a full doc audit.
Read an existing research outline and fan out independent background agents to deeply research each item, producing one structured JSON per item against the shared field schema. Resumable, batched, with output disabled per agent (each agent has its explicit output file). Use when `outline.yaml` + `fields.yaml` from `/research-outline` are in place and you want to parallelise the per-item investigation — academic paper deep-dives, product comparisons, benchmark fills, multi-target competitive analysis. Trigger this after `/research-outline`, before `/research-report`.
Power-user audit of Origin's pending surfaces. Most users want `/brief` for revisions. That handles the daily flow. Use `/review` only for explicit deep-walk audits after bulk imports, or when you want to walk the full queue rather than the top 3 shown in /brief. Invoked as `/review captures` or `/review revisions`.
Thoroughly improve code quality by requesting a review from a sub-agent after completing tasks requiring code editing, before returning the conversation turn to the user. Use for any task that edits code or configuration. Read this skill before starting edits, and after completing the requested changes run an isolated sub-agent review loop before returning the final response.
Panel review of any chosen scope: a review brief, parallel reviewer model voices (presets standard/max/ultra), merged findings, triage, and optionally fixes and verification. Fixes are user-gated by default; an upfront auto-fix option approves them. Spawns several fresh agent sessions, so cost scales with the preset; strongest with multiple harnesses installed (claude, codex, cursor), falls back to one.
Autonomous browser control via Chrome DevTools Protocol and Accessibility Tree for full screen control like GPT Operator. Use when user requests "control my screen", "verify Lovable", "test scrapers", "debug DOM", "autonomous testing", or any browser automation task. NO screenshots - uses CDP + semantic queries for 10x faster, 100% reliable element targeting.
SDD step 7. Archive feature.md and plan.md into docs/specs-archive/<yyyymmddHHMM>-<feature-name>/ directory, then update docs/project.md with the new feature, and any architecture decisions made. Use after /sdd-review is complete and the feature is ready to merge.
Configure the Serena MCP server — its global defaults and a repo's per-project settings. Use when the user says "configure serena", "set up serena for this repo", "serena project config", "serena global config", "tune serena for this codebase", "edit serena_config.yml", "edit project.yml", "monorepo serena", "serena picked the wrong language", "serena is indexing vendor/node_modules", "make serena read-only here", "serena contexts and modes", "what context for claude-code", "disable serena memory/onboarding", "point serena at my installed language server", "ls_specific_settings", "ls_path", "SERENA_HOME", or "move the .serena folder". Routes per-repo `.serena/project.yml` questions to `references/project-config.md` and machine-wide `~/.serena/serena_config.yml` (settings, contexts, modes, language-server tuning) to `references/global-config.md`. Do NOT use for general "use serena to find / edit X" routing — that is an MCP-routing concern, not configuration.
Set up and connect a new Coval agent for evaluation. Guides through agent type selection, endpoint configuration, system prompt, and default resource attachment. Use when user says "set up an agent", "connect my agent", "create an agent", "add an agent", or "configure agent endpoint".
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.