Skills
97,725 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Bootstrap the Roblox skill pack in a game project by injecting a `## Roblox Skills` block into CLAUDE.md — the full game-development lifecycle map, every skill's purpose, use case, and trigger words, and the next-skill routing so each skill ends by telling the user what to run next. Also records the project's tracker (GitHub issues or local roadmap), triage labels, and UI design artifact layout. Run once when setting up a Roblox project, or if the skills seem to be missing lifecycle, tracker, label, or routing context.
Digest of what your Superset agents did — sweeps workspaces, tasks, and agent terminals, then reports what finished, what needs review, and what's blocked. Use when the user asks what their agents did, wants a standup or summary of agent work, or returns after being away.
Renders planned changes — architecture and before/after comparisons, risk heat maps, execution order, dependency graphs, impact metrics — in your chosen output format (ASCII + emojis, an interactive HTML playground, or a NotebookLM infographic). Stores visualizations in memory for cross-session reference. Use when reviewing implementation plans, comparing approaches, assessing risk, or analyzing change propagation.
Manage `cs` (claude-statusbar) — switch theme/style/density, override severity colors, preview combinations, run doctor, reset config, install, upgrade (`cs upgrade` — the only supported upgrade path), or remove the bar, toggle fast/daemon mode, show cost or prompt-cache age, or toggle the activity segments (todos, active tool, running subagents, session duration, lines changed, git ahead/behind). Use whenever the user mentions cs, claude-statusbar, status bar, status line, 状態栏, 主题, theme switching, style switching, color customization, 余量颜色, 警告颜色, severity color, /statusbar, cs preview, cs doctor, fast mode, daemon, refreshInterval, 5h/7d window, context window display, prompt cache, todos / 待办, active tool, subagents / 子agent, session duration / 时长, lines changed / 行数, git ahead-behind / 领先落后, forecast / 预测 / 还能用多久, at-risk chip, show_forecast, or asks to install / upgrade / update / 升级 / configure / diagnose / customize the bottom status line in Claude Code.
Gradually form a revisable user profile over the conversation. Use when the user asks to create, view or update a profile, wants answers to fit them better, or repeatedly corrects "that's not like me", "too formal", "you misunderstood". Also use when the user says "remember my preferences", "default to this from now on", "follow my usual habits", "don't do that again", "don't record this", "delete the profile", "what have you noted about me", "you forgot", or in ordinary tasks that need to carry over known preferences. It only triggers when "remember" points to a long-term preference; in-task instructions like "remember this port number" don't count. Only record preferences, judgment habits, expression boundaries and open questions that affect later answers.
Review a Confluent agent skill in this repo against the Agent Skills spec (agentskills.io), Confluent conventions in CLAUDE.md, the PR template gates, and the evals-as-contract rule. Use this skill whenever the user asks to review, audit, validate, or lint a skill; opens or inspects a PR that adds or modifies anything under `skills/`; asks about spec conformance, lazy-loading, frontmatter shape, trigger overlap, or eval coverage; or wants a pre-merge sanity check on skill changes. Do NOT trigger for general code review of application code; security review; auditing schemas, producer/consumer configs, PII tagging, or Terraform generation for Schema Registry (handled by `kafka-schema-registry`); runtime/log analysis of skill behavior (use `tools/skill_review_dashboard.py`); or any changes that don't touch the `skills/` tree.
Author, scaffold, and run coding-agent-driven tests: Markdown case files executed by a coding agent against a live environment (browser, API, DB, logs, cloud, telemetry) with an auditable PASS/FAIL/BLOCKED report. Use when a deterministic test would be premature, brittle, too expensive, or too narrow.
Use when user requests deep analysis, thorough thinking, or detailed breakdown of a problem. Triggered by phrases like: 帮我深入思考, 请仔细分析, 帮我详细拆解, 请梳理一下思路, 仔细考虑, 深入理解, 详细分析, or similar expressions indicating need for systematic thinking. The skill applies ReAct-Plan framework: interleaving chain-of-thought reasoning with explicit global planning, dynamic prediction, and reflection to overcome myopic behaviors.
Используй, когда просят спроектировать систему, компонент или фичу и доступны тулы Искрона: «спроектируй», «давай продумаем», «архитектура», «нужен дизайн», design, plan the system, «жц», «крия», «эстафета», «записать в Искрон», «положить в граф». Также в моменты решений («мы решили», persist this design) и после брейнсторма — граф превращает дизайн в запись. Не для имплементации уже принятой конкретной спеки, рутинных фиксов и кодинга без открытого структурного выбора.
Build the project's shared language and bounded contexts before writing code, so names stay consistent and the agent stops paraphrasing domain concepts. Produces a CONTEXT.md glossary and decision records. Use at the start of a project or feature, or when the codebase and the people describing it speak different languages.
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
Drill open decisions one at a time — present the most important decision not yet clarified, give the top four choices, state a preference, ask the user. Use when the user says "next decision", "one decision at a time", or a plan has several unresolved choices. Differentiator: forward-looking; the decisions skill is retrospective (choices already made). Can be invoked with /next-decision.
Convert a local AGENT.md into a Claude Code optimized agent — an AUDIT-then-rewrite that ties every change to real Claude agent-runtime behavior, not taste. Inventories the agent, checks it against the runtime reference (the whole body becomes the system prompt; role identity stays, procedure moves to skills), writes a per-agent DAG rewrite plan with source-backed guardrails, and — only when asked — rewrites the frontmatter and prompt body thinner and more role-specific. Defaults to PLAN mode, never a blind rewrite, uses agent-native frontmatter only, and grounds each decision in the runtime source rather than inventing conventions. Use to migrate or thin an AGENT.md for Claude's agent runtime.
Turn a recurring chore into a Superset automation — drafts the agent prompt, confirms schedule and target, creates it with the CLI, and reviews the first run together. Use when the user wants a scheduled or recurring agent, a daily/weekly job, or to automate a repeating task with Superset.
Set up (or toggle) a macOS desktop banner that appears whenever Claude Code finishes a task, showing a one-line summary of what was done. Use when the user says "notify me when Claude is done", "desktop notification on completion", "set up the task banner", "mute the banner", or "unmute the banner".
Developer utilities and tools — CLI generation, skill scaffolding, agent delegation, system upgrades, evals, documents, parsing, audio editing, Fabric patterns, Cloudflare infrastructure, browser automation, meta-prompting, and aphorisms. USE WHEN create CLI, build CLI, command-line tool, wrap API, add command, upgrade tier, TypeScript CLI, create skill, new skill, scaffold skill, validate skill, update skill, fix skill structure, canonicalize skill, parallel execution, agent teams, delegate, workstreams, swarm, upgrade, improve system, system upgrade, check Anthropic, algorithm upgrade, mine reflections, find sources, research upgrade, PAI upgrade, eval, evaluate, test agent, benchmark, verify behavior, regression test, capability test, run eval, compare models, compare prompts, create judge, view results, document, process file, create document, convert format, extract text, PDF, DOCX, XLSX, PPTX, Word, Excel, spreadsheet, PowerPoint, presentation, slides, consulting report, large PDF, merge PDF, fill form, tracked changes, redlining, parse, extract, URL, transcript, entities, JSON, batch, YouTube, article, newsletter, Twitter, browser extension, collision detection, detect content type, extract article, extract newsletter, extract YouTube, extract PDF, parse content, clean audio, edit audio, remove filler words, clean podcast, remove ums, cut dead air, polish audio, transcribe, analyze audio, audio pipeline, fabric, fabric pattern, run fabric, update patterns, sync fabric, summarize, threat model pattern, Cloudflare, worker, deploy, Pages, MCP server, wrangler, DNS, KV, R2, D1, Vectorize, browser, screenshot, debug web, verify UI, troubleshoot frontend, automate browser, browse website, review stories, run stories, web automation, meta-prompting, template generation, prompt optimization, programmatic prompt, render template, validate template, prompt engineering, aphorism, quote, saying, find quote, research thinker, newsletter quotes, add aphorism, search aphorisms.
Plan a feature or task in fast, full, or ultra mode. Ultra creates an indexed multi-file bundle with deeply specified phases for execution by a smaller model. Use for "plan", "new feature", "start feature", "create tasks", or exhaustive implementation planning.
Use this whenever building AI SDK agents, chat apps, tool-calling workflows, human-in-the-loop systems, streaming UIs, artifact generation, evaluator-optimizer loops, multi-agent orchestration, structured outputs, or production agent patterns. This skill distills architecture and implementation lessons from the aisdkagents.com pro block library so an AI assistant can recommend proven patterns and explain how to build them.
Exact-name Codex/OMX-compatible $autopilot front door for autonomous idea-to-verified-code work. Use when the user invokes $autopilot, autopilot, auto pilot, full-auto, or asks for an end-to-end build; route small focused edits to direct implementation and stricter spec-first loops to ooo or ralph.
Use this skill when an LLM should act inside a business process — choosing tools or next steps, calling systems, handing off to people — with guardrails, human review, or an audit trail, even if Camunda isn't mentioned, and before hand-coding an agent loop. Models Camunda 8 AI Agent sub-processes: tools as BPMN tasks, fromAi() parameters, prompts, and limits. Also covers v1/v2 element-template selection and upgrading v1 elements to v2.
Give the human a fast, plain-English catch-up on what changed in the project: what the agents did, why, and what decisions need their input. Use when the user asks to "catch me up", "what changed", "where are we", "recap", "brief me", "give me the rundown", "what did you do", "summarize the session", or "fill me in", or otherwise wants to get back up to speed after being away.
End-to-end content creation workflow that orchestrates research, editorial review, and social distribution agents in sequence. Use when the user wants to create, review, and distribute content through a multi-stage pipeline, or says "/content-pipeline".
Use when anyone (user or agent) is about to author any structured content — RFC, ADR, runbook, spec, plan, PRD, README, retrospective, blog post, PR description, issue body, ticket, multi-paragraph commit message, announcement, or any other shape — regardless of destination (local file, GitHub, Jira, Linear, Slack, email, Confluence, Notion, Google Drive, etc.). This skill is the gate every content-authoring action passes through. It (1) identifies content-type + variant from intent via research, (2) checks docs/templates/<content-type>/<variant>/ for an existing template, (3) if it exists, returns it plus a hard-refusal directive that you MUST follow, (4) if missing, forges a new one via research, then returns it under the same directive, (5) on "advise:" intent, helps place content that doesn't fit (maps to sections or proposes new variant). Modes: use-or-create (default), force-regenerate, force-new-variant. Self-contained: prefers a research-capable skill if available, else built-in WebSearch + WebFetch.
Routing front door for a structured, human-in-the-loop deep-research workflow (Weizhena/Deep-Research-skills) — turn a topic into an extensible outline, fan out parallel web-search agents to investigate each item into validated JSON, then render a complete markdown report. One skill, 4 reference pipelines: outline (research / add-items / add-fields), deep (parallel per-item investigation + field-coverage validation), report (TOC + per-field markdown), web-search (research agent + 5 routed source modules). Plugin: npx skills add https://github.com/akillness/jeo-skills --skill deep-research
Model-routing orchestration for any expensive frontier model (Fable, Opus, GPT-5.x) - the main model keeps judgment and Q&A review, mechanical subagent dispatches carry an explicit cheaper model, and ax measures whether the routing actually worked. Use when orchestrating codebase-heavy or token-heavy work with subagents, when dispatching Agent tasks without a model, when the user says "route to cheaper models", "efficient dispatch", "optimize model spend", or asks where their token spend goes. Pairs with the route-dispatch hook (deterministic backstop) and `ax dispatches` (evidence). Do NOT fire on single-shot questions or tiny tasks with no dispatching.
Relentlessly interrogate the user about a plan or design until shared understanding is reached, resolving every branch of the decision tree. Use when the user wants to stress-test a plan, be challenged on their design, or mentions "grill me" / "interroge-moi" / "challenge-moi" / "questionne-moi".
Add new/custom AI models to opencode.json. Use proactively for bleeding-edge releases, non-standard providers, self-hosted models, or custom endpoints. Examples: - user: "Add the new Claude 4.5" → websearch for API specs, add provider entry with baseUrl, verify model ID format - user: "Use my local Ollama instance" → configure custom provider with http://localhost:11434, set model ID format - user: "Configure this OpenAI-compatible proxy" → add provider with custom baseUrl, set apikey env var, verify compatibility - user: "Model X just released, add it" → research provider documentation, find model ID and capabilities, add to config
Orchestrate one-click generation of major and detailed plot points: preprocessing, extraction, refinement, and formatting. Use when a complete plot outline with point-by-point development is needed, or when a plot workflow or plot-point pipeline is mentioned.
The index of every pro-workflow skill and command, grouped by job, with when to reach for each and whether it is human-run or auto-triggered. Use when you are not sure which skill fits, want the full map, or ask "what can this do", "which skill for X", "list the workflow".
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.