Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Persistent, recoverable GSAP workflow for coding agents. Use when the user wants to animate a new page or section, or improve an existing page/component with GSAP. Route new/greenfield work to gsap-new and existing-code improvement work to gsap-refactor. Always use .gsap artifacts as the source of truth so any fresh agent session can resume without chat history.
Use when working with Harbor's `harbor exec` CLI workflow: compiling files, directories, or globs into Harbor tasks; running map jobs; configuring artifacts and existence-only verification; using map-reduce; writing or reviewing ExecConfig YAML/JSON/TOML; or debugging command behavior, config validation, and job outputs.
Install and configure the OpenViking long-term memory plugin for OpenClaw via natural conversation. Once installed, the plugin automatically captures facts from chats and recalls relevant context before each reply (auto-capture + auto-recall, cross-session). Covers prerequisites, install through OpenClaw's plugin manager (ov-install as backup), wizard-based configuration, slot activation, gateway restart, verification, multi-tenant root keys, multi-instance and uninstall. The user needs no CLI knowledge — the agent runs everything and only asks for a few values. Assumes the OpenViking server is already running; it does NOT install the server. Trigger on: "install OpenViking", "set up memory", "configure memory plugin", "add long-term memory", "semantic memory", "RAG", "帮我装 OpenViking", "配置记忆插件", "安装记忆功能", "接入 OpenViking".
Run MassGen experiments and analyze logs using automation mode, logfire tracing, and SQL queries. Use this skill for performance analysis, debugging agent behavior, evaluating coordination patterns, and improving the logging structure, or whenever an ANALYSIS_REPORT.md is needed in a log directory.
Proven workflow architectural patterns from real n8n workflows. Use when building new workflows, designing workflow structure, choosing workflow patterns, planning workflow architecture, or asking about webhook processing, HTTP API integration, database operations, AI agent workflows, batch processing, or scheduled tasks. Always consult this skill when the user asks to create, build, or design an n8n workflow, automate a process, or connect services — even if they don't explicitly mention 'patterns'. Covers webhook, API, database, AI, batch processing, and scheduled automation architectures. Also use when optimizing a slow workflow or speeding up large-item-count processing (node count, batchSize, all-items vs per-item).
Surface the Opik traces worth a developer's attention, ranked by signal — Diagnostics issues first, then errors, failed tool calls, latency, regressions, and low online-eval scores. With the Opik MCP connected it lists the project's agent_insights_issue entities, offers to turn Diagnostics on when the project has it switched off, then fills the gaps with list (filters, sort, a time window); without the MCP it reads the same via the SDK (agent_insights and search_traces), so it works with no MCP. Returns a ranked shortlist, each item ready to hand to the explain skill. Use for "what is broken in production", "which traces need attention", "find failing or slow traces", "which tool calls are failing", "triage my agent". Not for offline experiment results (use evaluate or compare) and not for root-causing one trace (use explain).
Root-cause a specific Opik trace, or a pattern across traces, and return a grounded explanation. Uses the hosted Opik MCP when it is connected, and falls back to SDK scripting otherwise. Returns the root cause, the evidence spans as clickable Opik UI links, and one suggested next step. Use for "why did this trace fail", "explain this trace", "debug this trace", "why is my agent slow or wrong". Not for adding tracing to an app (use the instrument skill) or for changing code.
Dependency-graph-style task breakdown on PLAN.md — answers 5 questions (what does this task need · what runs in parallel · what does it depend on · why does it exist in the graph · how does it relate to the old graph) with deterministic tools, dispatch by topo layer with LOCK + lease + generation, state persisted append-only (events.jsonl), 2 Python files for drawing (1 graph / 2D atlas of all graphs), every model answer labelled sure|suggestion|unknown + source, fabrication audit = 0. Invoke when the user says "orca-graph", "graph phân việc", "task nào song song", "phụ thuộc gì", "vẽ graph task", "dispatch theo graph", "/orca-graph".
Apply to any non-trivial work, not just bulk work: edits, migrations, analyses, checks. Build the tool that does it or proves it (codemod, script, generator, or a skill your subagents follow) instead of working by hand. The tool is the artifact a reviewer can rerun
Evaluate how well a codebase supports autonomous AI development. Analyzes repositories across eight technical pillars (Style & Validation, Build System, Testing, Documentation, Dev Environment, Debugging & Observability, Security, Task Discovery) and five maturity levels. Use when users request `/readiness-report` or want to assess agent readiness, codebase maturity, or identify gaps preventing effective AI-assisted development.
Configure which models pstack uses per role and at what reasoning budget. Detects your available models and writes ~/.config/pstack/pstack-models.md, which overrides the skill defaults. Use for /setup-pstack, "configure pstack models", "pstack budget", or changing pstack's model choices
Skill Hub - a Chinese platform for AI skill search, installation and smart recommendations. Use when users need to search for, discover or install Skills. Automatically recommends suitable skills when users lack a capability (e.g. needing images after writing an article, visualization after analysis, or deployment after coding). Triggers when users say "help me find a skill for XX", "is there a Skill for XX", "I need XX functionality", "install skill", "search skill", "find skill", "install skill". Keywords: Skill Hub, skill search, skill installation, skill marketplace, skill recommendation, capability gap, skill marketplace, find a skill, install a skill, skill hub
Applies the user's engineering standards during substantial technical work: preserve scope, use current evidence, delegate independent parts, and complete authorized outcomes. Use during substantial implementation, debugging, refactoring, or multi-part delivery work.
Decompose high-level objectives into atomic implementation tasks for Python/React projects. Use when breaking down large features, multi-file changes, or tasks requiring more than 3 steps. Produces independently-verifiable tasks with done-conditions, file paths, complexity estimates, and explicit ordering. Creates persistent task files (task_plan.md, progress.md) to track state across context windows. Does NOT cover high-level planning (use project-planner) or architecture decisions (use system-architecture).
Patterns for delegating UI work to agents: one synthesis agent versus specialized agents, multi-agent pipelines for research, design, implementation, and review, and clean handoffs between them. Use when a UI task is too large for one pass or you are designing an agent workflow for components or design systems.
Operates the git-agent CLI — atomic commits, init/config, and the code graph (event log + AST + co-change). Use it whenever the user wants to commit or set up git-agent; when you are about to modify a feature and need the files that move with it (graph impact); when you are changing a function and want its callers/callees (graph callers/callees); when you need a symbol's location or source (graph node/query); when deciding which tests to run after a change (graph affected); when a test broke and you want to trace the action that introduced it (graph diagnose); or when you need action history (graph timeline) or a file's rename-aware provenance (graph provenance). All graph queries are read-only and offline (no LLM, no API key); only commit and init --scope need a provider.
All online operations must be handled through this skill, including searching, web scraping, post-login actions, and online interactions. Use when the user asks to search for information, view web content, visit a site that requires login, interact with a web interface, scrape social media content (Xiaohongshu, Weibo, Twitter, etc.), read dynamically rendered pages, or perform any online task requiring a real browser environment.
End-to-end ticket implementation with a multi-agent team using TDD. Retrieves JIRA ticket details and Figma designs, spawns a PO to write requirements, gets developer approval, then proposes a test plan for review. Once tests are approved, implements tests first (red), then production code to pass them (green). Use --no-jira to skip JIRA retrieval and use inline instructions instead. Use --headless for fully autonomous execution where a reviewer agent replaces human approval.
Use when users explicitly ask to brainstorm, shape requirements, compare approaches, or produce a planning/design document before implementation. Turn rough ideas into validated brainstorming/design docs through focused dialogue, role-based analysis, and explicit user approval. Do not trigger this skill for straightforward implementation requests that do not need dedicated design exploration.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Use when the user asks "what can Cekura do", "what commands are available", "help me with Cekura", "what skills do I have", "show me Cekura features", "what's available", "how do I use Cekura", or needs guidance on which Cekura skill to use for their task. Also relevant as the entry point when a user has just installed cekura-skills for the first time.
Adversarial pre-execution gate. Activates before any plan, proposal, or side-effecting action (create, edit, delete, run, deploy, migrate, call a tool/MCP/agent, git write) and before architecture, product, UX, data, vendor, or strategy decisions. Scales its depth to the real risk of the action — a one-line note for trivial work, a short evidence-based critique for normal changes, a hard stop with targeted questions for critical ones — and ends with a gate that waits for the user's explicit approval before acting. Read-only requests and trivial, instantly reversible edits pass with at most one line. Use whenever the user shares a plan, asks "is this a good idea", "review this", "what could go wrong", or asks to execute a change.
Use when you need to visually interact with a GUI: test buttons, fill forms, verify visual layouts, fuzz web pages, automate user flows, take screenshots, or perform end-to-end QA on any application. Works on cloud VMs, Docker containers, local machines, and sandboxes. Needs the `cua` CLI (curl -fsSL https://cua.ai/install.sh | sh).
Use when a session is running out of context and the work needs to continue in a fresh one, or when starting a session meant to pick up where an earlier one stopped. Not for compacting in place — that is /compact — and not for specifying work that has not started, which is a plan.
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.