Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Exports the project's Claude configuration (CLAUDE.md + ai-context/) to tool-specific instruction files for GitHub Copilot, Google Gemini, and Cursor. Trigger: /config-export, export config, copilot instructions, gemini config, cursor rules.
CoPaw operations assistant. Use when the user mentions copaw operations, unresponsive services, disconnected channels, MCP failures, model call failures, cron not running, Docker deployment, reload, restart, or reset recovery. Prioritize status checks and fault triage; for high-impact actions such as restart, reload, reset, or configuration changes, explain to the user before executing.
CTF general solving orchestrator. Use when facing a CTF challenge of unknown type and needing to automatically analyze the challenge type and choose the correct solving path. Automatically dispatches the corresponding specialized skill (pwn/crypto/web/reverse/forensics/osint/malware/misc); suited to scenarios where a challenge file or service endpoint is given but the category is uncertain
Migrate Dialogflow CX (DFCX) agents to CXAS (Customer Experience Agent Studio) agents. Use this skill when the user mentions DFCX migration, migrating agents, converting DFCX to CXAS, porting agents, agent migration, or post-migration optimization/consolidation. Four independently runnable scripts: migrate.py (1:1), stage_1.py (variable dedup + consolidation), stage_2.py (instruction state machines + tool mocks + lint + report), stage_3.py (rewires consolidated topology from source dep graph; only needed when stage_1 ran consolidation). State persists between scripts via <target>_ir.json so each can run / re-run / resume independently.
Comprehensive adversarial audit of a theory, proof, math/econ paper, codebase, or set of claims — decompose into components, fan out independent skeptics that must return CONCRETE defects, adjudicate every finding with a separate judge, fix all confirmed defects, then re-verify. Use when correctness must be bulletproof and single-pass or round-by-round review is too slow and too shallow. Invoke for "audit this rigorously", "find ALL the bugs/gaps", "make this rock solid", "converge faster on correctness".
Use this skill whenever the user wants to call DeepSeek V4-Pro / V4-Flash (or its legacy aliases deepseek-chat / deepseek-reasoner), or you see code that imports `from openai import OpenAI` with `base_url="https://api.deepseek.com"`. This skill teaches you the 10 protocol contract rules required to avoid the 16 documented bugs (reasoning_content lifecycle 400, tool-call leakage, max_tokens runaway → V8 string limit, parallel tool delta interleave, prefix-cache invalidation, /beta endpoint remap, etc.) and ships a single drop-in `safe_init.py` script you can copy into the user's project.
Hard gate before production implementation: do not write behavior/production code until the chosen solution/design is approved; named escapes for auto mode, Jira auto/force, and lean hotfix require short 留痕. Referenced by PDCA hosts before execution stages. Triggers — 「设计批准门禁」「方案批准后才能写码」「实现前批准」「design approval gate」 / design approval gate, pre-impl approval. Do NOT use as a name alias for brainstorming.
Complete 9-phase development pipeline guide — from schema to deployment. Pipeline phases (1-schema → 9-deployment) are orthogonal to PDCA's 9-phase per-feature cycle and Sprint's 8-phase container; each pipeline phase may host PDCA cycles for individual features, and multi-feature pipeline initiatives can be wrapped in /sprint (v2.1.13). Triggers: development pipeline, where to start, phase
Use for deep design discovery and problem understanding before planning or implementation. Read the principles skill first. Trigger when the user wants to think deeply, build knowledge, find edge cases, ask many questions, or explore the problem space without producing plans or code; save a design artifact when the user asks for one or future work needs it.
Maintain an append-only, no-rewrite project document/log history mechanism: for each change, create YYYY-MM-DD_主题.md and link them together with an INDEX.md timeline index, so agents such as CodeBuddy / Claude Code / Codex can read the full history. Use when the user says "记录一下", "写个日志", "追加文档", "整理文档历史", "建个 docs 索引", or "把这次改动记下来". model-invoked.
Use first when Arduino or embedded hardware work spans sessions, parts delivery, wiring, soldering, flashing, measurements, field tests, recovery, or a durable next-todo is needed. Maintain explicit evidence and concrete physical-world user gates instead of assuming completion.
[Beta] CI-only eval regression runner using gh-aw (GitHub Agentic Workflows). Runs all eval cases in .evals/ on a schedule or per-PR, reports pass/fail results, and can block merges on regressions. Also creates new eval cases from promoted patterns flagged by learning-aggregator-ci. Use when: you want automated regression testing of promoted rules in CI/headless pipelines. For interactive eval creation and runs, use eval-creator.
Operate web pages with a local model through the fbu CLI on Linux, Windows or macOS (PyTorch CUDA/CPU or Apple Silicon MLX). Use for browser navigation, searches, standard forms and dropdowns when the user wants local inference; accept any starting URL and natural-language goal.
An always-available field lab for thinking with AI, guided by Kit. Use it for any question, from a factual query or practical problem to a genuine tension, hostile thesis test, high-stakes decision, or full recursive dialectic. Give direct answers when enough. Treat open-ended requests to understand, explain, or make sense of conceptual or interpretive material as caddying prompts, not permission for a long explanation: recommend a concrete way to examine the material and run only what the user selects. For other nontrivial inquiries, ask what the user hopes to accomplish when that would change the instrument. Return only what each operation supports; synthesize, recommend, decide, plan, or act only when asked. Offer a Field Log when sources, findings, and open questions need to stay together; collect related logs in an Expedition; use Essay to find and develop source-grounded essays from completed Field Logs; run the Electric Monk dialectic only as a selected workflow.
Expert guide for Glove primitives and packages such as glove-core, glove-facts, dynamic goals, form preparation, glove-react, glove-next, tools, display, models, stores, memory, voice, image, and working environments. For full agent applications, runtime orchestration, or migration from deprecated Glovebox, use the glove-foundry skill.
Prepare a read-only structured goal of at most 4000 characters as an assignment for an LLM agent, not a human-facing work item. Use when the next executor is an LLM agent and the assignment must be a verifiable read-only goal, not a human-facing work item. Russian discovery terms: проверяемая цель.
Manage hooks and automation for coding agents (Claude Code, Codex CLI, OpenCode, Devin CLI/Desktop). Use when users want to add, list, remove, update, or validate hooks. Triggers on requests like "add a hook", "create a hook that...", "list my hooks", "remove the hook", "validate hooks", or any mention of automating agent behavior with shell commands or plugins.
Steel-manning resolution for tempa-spec's dark mode. Resolves findings that would normally escalate to the user (DECISION/HIGH from hrt-align-consistency-review and hrt-apply-code-review; single-reviewer UNANCHORED from hrt-adversarial-authoring). Always a fresh subagent. Use only when dark.md exists and the calling skill has a finding it cannot resolve on its own evidence.
Opt-in gate for tempa-spec's dark mode (bounded autonomous decision-making across discovery-through-align, optionally through apply). Never offered proactively - only activates when the user asks for it unprompted, in their own words. Shows a mandatory risk warning before any confirmation, then asks which variant (planning-only or +implementation) and creates dark.md. Use once, at or after discovery, only on explicit user request.
Single source of truth for where an escalation-worthy finding goes when tempa-spec's dark mode is active (dark.md exists). Each of proposal, specs, design, align, and apply points here instead of inlining its own dark-mode carve-out. Says which findings route to hrt-dark-mode-decision-gate, which still go to the user, and which dark.md section records the result.
Convert SPEC.md to spec.json, craft the implementation prompt, and execute the iteration loop via subprocess. Use when converting specs to spec.json, preparing implementation artifacts, running the iteration loop, or implementing features autonomously. Triggers: implement, spec.json, convert spec, implementation prompt, execute implementation, run implementation.
Transform infrastructure documentation, runbooks, and operational knowledge into reusable Claude Code skills. Convert Proxmox configs, Docker setups, Kubernetes deployments, and cloud infrastructure patterns into structured, actionable skills.
Behavioral guardrails to prevent common LLM coding mistakes — enforces surgical changes, assumption verification, and scope discipline before and during implementation. Use when implementing features, modifying code, or when you notice yourself about to make changes without reading the existing code first.
Driving a code change or document update through reviewed research, plan, and design outputs → approval → execution → verification → independent review, including blocked verification, pending review/CI, and run recovery. Owns phase placement, approval revision, evidence, finding state, and recovery meaning—not generic deliverable-review method, check-failure diagnosis, host execution APIs, or delivery storage. Development practices own verification diagnosis; conduct practices own non-code deliverable review; project policy chooses gates and branches. Defer to a host project's own more-specific change loop where present. Not for standalone read-only questions, investigations, or reviews that change nothing.
Migrate configuration from Bluejay voice AI testing platform to Coval. Use when customer says "migrate from bluejay", "bluejay migration", "import bluejay config", or needs to transfer agents, simulations, metrics, and schedules from Bluejay to Coval.
24/7 autonomous business operations orchestrator with business health scoring, canary monitoring, and safety guardrails. Schedules and runs all business functions automatically — content publishing, social media posting, outreach sequences, ad monitoring, financial reporting, health checks, and post-deploy verification. Use when the user wants to automate operations, schedule tasks, set up autonomous workflows, or says 'automate this', '24/7', 'run automatically', 'schedule', 'cron', 'autonomous', or 'set and forget'.
Append todos, logs, learnings, and reflections to Obsidian daily or weekly notes, and run the agent's own #agent-todo queue, via scripts/capture.py. Use when asked to add a todo, note something down, log this, capture an idea, set a due date, record a weekly reflection, or list/close the agent's assigned tasks ("看你的 todo", "干活", "what's on your todo list"). Do NOT use for searching notes (use obsidian-search) or for creating standalone long-form notes.
DEPRECATED, renamed to orchestrating-chat-agents. Kept so existing imports of its scripts keep working; every module here loads the same module from orchestrating-chat-agents. Use orchestrating-chat-agents for parallel Claude calls over the API from claude.ai, and dispatching-subagents for fan-out in Claude Code.
ALWAYS apply by default in every turn when writing to the user, naming code/files/folders/APIs/terms, explaining decisions, summarizing work, or teaching concepts. Hard-enforces ASD-STE100 Simplified Technical English for all agent prose (approved simple words, one word per idea, short active sentences, short paragraphs) plus readable names. Also use when the user says plain-language, STE, Simplified Technical English, simpler words, explain simply, less jargon, better names, or readable naming.
Self-evaluate a plan on a 100-point scale after it is created or updated. Make sure to use this skill immediately whenever you create a plan or update a plan, even if the user does not explicitly ask for a review. This skill ensures that the plan is clear, comprehensive, feasible, and consistent before execution.
Project initialization tool. Reads the global protocol ~/.claude/CLAUDE.md, analyzes the project's actual situation, and generates project-specific CLAUDE.md and docs/ context. This skill should be used when the user says "initialize project", "project setup", "configure Claude Code", "new project configuration", or when entering a new project that needs quick configuration. Do not use for: Skill content development or acceptance (use skill-lint), one-off Skill installation (use skill-manager), or code generation.
Use when managing prompt templates, running evaluations, building workflows, logging LLM requests, or analyzing AI application performance. Reach for this skill when working with prompt versioning, release management, observability, tables/datasets, or multi-step AI workflows.
Convert PRDs to prd.json format for the Ralph autonomous agent system. Use when you have an existing PRD and need to convert it to Ralph's JSON format. Triggers on: convert this prd, turn this into ralph format, create prd.json from this, ralph json.
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.