Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Writing or editing a Ring skill: SKILL.md structure, frontmatter and Agent-Search-Optimization rules, token-efficiency targets, and bulletproofing (Iron Law, rationalization tables, Red Flags) so discipline-enforcing skills resist excuses. Use when creating or revising a skill. Delegates pressure-testing to ring:testing-skills-with-subagents. Skip for pure reference skills with no rules, or project conventions (put those in CLAUDE.md).
Use when asked to audit, measure, or re-check how agent sessions are using the Bash tool — command chaining (&&, ;, |), cd into the session's own repo, head/tail truncation, sed -n/cat/heredoc instead of Read/Edit, git commit/push inside chains or behind git -C — or when deciding whether a permission prompt, an allowlist rule, a ~/.agents/AGENTS.md Bash rule, or the permission mode (acceptEdits vs auto) needs changing and wants evidence from real transcripts rather than a hunch. Runs a stdlib script over ~/.claude/projects/*.jsonl, prints per-model and per-session rates plus samples, and carries the dated research that explains why each pattern happens and where the fix belongs. Also the place to record a newly noticed Bash anti-pattern so the next audit measures it.
View and configure settings for coding agents (Claude Code, Codex CLI, OpenCode, Devin CLI/Desktop, and others). Covers JSON settings for Claude Code and Devin, TOML for Codex CLI, and JSON/JSONC for OpenCode, including permissions, sandbox, model selection, profiles, feature flags, providers, hooks, subagents, and skills.
Set up git guardrails that make coding agents ask or stop before commands that lose work or rewrite shared history, at a guard level the user picks, for whichever agent harnesses they use. Use when the user wants to guard against destructive git operations by coding agents.
Sets up an `## Agent skills` block in AGENTS.md/CLAUDE.md and `docs/agents/` so the engineering skills know this repo's issue tracker (GitHub or local markdown), triage label vocabulary, and domain doc layout. Run before first use of `to-issues`, `to-prd`, `triage`, `diagnose`, `tdd`, `improve-codebase-architecture`, or `zoom-out` — or if those skills appear to be missing context about the issue tracker, triage labels, or domain docs.
Keep a reviewable decision trail for long-running or unattended work: a TSV log with one row per decision (what, why, evidence, result). Local by default; commit it when a reviewer needs the trail to trust the result. Use for /show-me-your-work, autonomous or multi-phase runs, or work a human reviews after stepping away.
Scaffold and formally review a new Agent Skill in this marketplace repo. Covers valid SKILL.md generation, directory layout, registry entries, and spec auditing. Use when adding a skill, validating frontmatter, or checking marketplace readiness before a PR.
Build correct, consistent Agent Skills for any compatible agent product. Use when creating, updating, renaming, deleting, or validating a skill, or when improving skill descriptions, structure, safety rules, progressive disclosure, or trigger reliability.
Create new Claude Code Skills with proper YAML frontmatter, progressive disclosure structure, and complete directory organization. Use when you need to build custom skills for specific workflows, generate skill templates, or understand the Claude Skills specification.
Improves Agent Skills via validate → rule-based eval cases → plugin-eval → prompt evals → bounded edits with held-out gates. Use when tuning skill quality, routing, or adopting Chrome/Microsoft T-named quality gates—not for bulk validate-only or SkillOpt automation.
Requires durable citations when authoring or researching Agent Skills—maintain references/sources.md per skill, link external research, and record provenance in PRs. Use when creating skills, updating SKILL.md, doing web research for skills, or auditing missing sources.
Tmux execution support for long-running and persistent commands. Load this skill when you need to: (1) run commands expected to exceed tool timeout (over 60s), (2) start persistent servers or dev processes, (3) run TUI/interactive applications, (4) execute parallel isolated tasks in separate sessions, (5) run remote commands over SSH that must survive disconnection, (6) handle tmux errors like 'capture-pane blocked in interactive_bash'. Trigger phrases: 'run in background', 'start server', 'long-running', 'tmux session', 'keep running', 'persistent process', 'dev server', 'training script', 'git clone large repo', 'docker build', 'capture-pane blocked'.
Administrative workflows for the agent-skills repository. Use when the user wants to contribute a skill, open a pull request, or update an already-installed skill to the latest version. Trigger when users say: "open a PR", "submit my changes", "push this skill", "update my skills", "update the skills repo", or "how do I contribute a skill".
Post-implementation completion workflow for Spec-backed Plans. Use after spec-implement completes to validate, review, create stacked commits, and open a PR via code-pull-request. Triggers only with an active Spec-backed Plan after spec-implement completes, including when the user says "finish", "done", or "complete" in that context.
Skills sync assistant. Sync skills from a local directory, a GitHub repository, or a skillsmp.com page into the AI coding tool directory chosen by the user. Use AskUserQuestion to pick targets (multi-select and custom paths supported), show the sync plan before executing; skills with the same name are overwritten. Must include ~/.agents/skills as a sync target.
Create high-quality, model-aware system prompts for any LLM (Claude, GPT, Gemini, open-source, etc.). Use this skill whenever the user wants to create, write, build, design, draft, or improve a system prompt, system instructions, or custom instructions for any AI model. Also trigger when the user asks about prompt engineering, prompt design, prompt optimization, or wants to define behavior for an AI assistant, chatbot, agent, or any LLM-powered application — even if they don't explicitly say "system prompt". Covers all use cases including chatbots, agentic systems, tool-use workflows, content generation, data extraction, code assistants, and multi-turn conversations.
Use this skill whenever you need to interact with tmux — sending commands to panes, reading their output, watching a TUI app like vim, htop, or lazygit, or managing sessions and windows. Reach for it any time the user says "run this in the terminal", "check that pane", "send a command to the shell", or when you spawn a background task that you'll need to monitor. Always use this skill when tmux is involved — it provides reliable pane identification, delta output reading, and safe key-sending that prevent common mistakes.
Use this skill when work needs durable planning, adaptive product slicing, discovery before delivery, resumable sprint execution, wave or initiative scaffolding, source-pack routing, closeout evidence, final audits, or repeatable product-agile workflow structure. Do not use it for small direct fixes that do not need a durable artifact, or for governance-only questions that belong to ub-governance.
Apply plugin knowledge base updates to an existing generated system. Consults the Ars Contexta research graph for methodology improvements, proposes skill upgrades with research justification. Never auto-implements. Triggers on "/upgrade", "upgrade skills", "check for improvements", "update methodology".
Build realtime voice / live audio agents with AG2's `ag2.live` module. Wrap a prompt + provider config in `LiveAgent` and open a bidirectional voice session with `agent.run()`, pumping mic audio in and playing synthesized speech out. Covers the two realtime providers — Gemini Live (`GeminiRealTimeConfig`) and OpenAI Realtime (`OpenAIRealTimeConfig`) with audio/text output modalities, voices, and user-speech transcription; audio I/O over the local sound card (`SoundDeviceRecorder` / `SoundDevicePlayer`, both sounddevice-backed); one-shot speech-to-text (`OpenAITranscriber`, `OpenAITranslationTranscriber`) and its `.pipe(agent)` voice pipeline; text-to-speech (`OpenAITTSConfig`); and `TTSObserver`, which speaks a regular text `Agent`'s streamed tokens aloud. Use when the user wants a talking agent, a phone/voice assistant, live transcription, or to add TTS playback to a text agent.
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to “open a website”, “fill out a form”, “click a button”, “take a screenshot”, “scrape data from a page”, “test this web app”, “login to a site”, “automate browser actions”, or any task requiring programmatic web interaction.
Inter-agent communication patterns including message passing, shared memory, blackboard systems, and event-driven architectures for LLM agentsUse when "agent communication, message passing, inter-agent, blackboard, agent events, multi-agent, communication, message-passing, events, coordination" mentioned.
Sets up and drives TencentDB Agent Memory (MemoryCore, MemoryHub, MemoryProxy, MemoryPanel) for agent teams. Use when the user mentions “memory hub”, “team memory”, “chat memory”, “skill library”, “wiki”, or “codegraph”, or asks to install, configure, or troubleshoot the memory-tencentdb plugin, onboard a new agent, or deploy standalone (Docker Compose, panel UI, team review).
Use when the request is already narrowed to wiring a model endpoint, env vars, settings-file-based model config, `${ENV.xxx}` placeholders, `auto_load_env=True`, or connectivity check for a model-powered feature, including local Ollama, dotenv-loaded DeepSeek or other OpenAI-compatible settings, plugin namespace placement, auth, request options, and minimal verification.
Use when the user is shaping how one model request or request family should be instructed or templated, including prompt slots, input/instruct/info layering, mappings, recursive placeholder injection, prompt config, YAML or config-file-driven prompt behavior, and reusable prompt structure.
Automake owns a Git-backed ratchet for safely improving a repository through separate optimizer and evaluator roles: it establishes a baseline, proposes and checks candidates, obtains an independent comparison against the current committed state, keeps accepted changes, and restores rejected ones exactly. It triggers when the request is to prepare or execute this bounded improvement workflow, draft or approve its role instructions, set limits or stop conditions, protect local files, initialize the baseline, or invoke `/automake`. It does not trigger when “automake” refers to the GNU build system, when the task is merely a Makefile target or normal bug fix, when the user only wants code review or an explanation, or when the request is generic branch reset or deleted-file restoration.
Work autonomously on a project without waiting for direction. On command, survey the codebase, pick one high-value low-risk improvement, do it, verify it, commit it, then pick the next, looping like a self-directed intern. Use when the user hands off open-endedly ("work on your own", "keep improving this", "find something worth doing", "don't wait for me", "go autonomous", "run for a while", "be a passionate intern", "do whatever you think is valuable") or invokes /autopilot, and you have to choose the work yourself rather than wait for instructions or feedback.
Use at the start of any bug bounty hunting session, when switching targets, or when feeling lost about what to do next. Master orchestrator combining the 5-phase non-linear hunting workflow with the critical thinking framework (developer psychology, anomaly detection, What-If experiments). Routes to all other skills based on the current hunting phase. Also use when asking “what should I do next” or “where am I in the process.”
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.