Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Enforces workflow discipline across ALL projects. Ensures Claude checks for specialized agents before responding, announces skill/agent usage, and creates TodoWrite todos for multi-step tasks. Triggers: automatic on all sessions, use the right agent, follow workflow.
Draft, execute, or track lean, verifiable implementation plans for non-trivial migrations, refactors, PR splits, task checklists, and other work with sequencing, tradeoffs, unknowns, or risks, and delete them when complete; use writing-roadmap for strategic direction and skip small obvious tasks.
Use to turn a spec, PRD, or epic ticket into the fewest dependency-ordered one-PR implementation tickets that ship safely, adopting and merging any hand-created tickets, applying the agent-ready body contract, configured estimates, and kind labels, and emitting a dependency graph and predicted file footprint.
Delegate real coding work to a peer AI agent (such as Claude Code, Codex (ChatGPT), or Gemini (Antigravity)) that runs its own full harness, or use that peer as an independent reviewer, adversarial red-team, ideation & planning partner. Trigger when the user says "use agent bridge", "ask <another agent>", "delegate to <agent>", "have <agent> implement/build/fix this", "get <agent> to review/critique", "get a second opinion", or otherwise hands work to a different agent than the one they're talking to. The bridge auto-detects which agent you are and offers the others as peers, so the same instructions work whichever agent you are running inside. Do NOT trigger for normal work the user wants you to do yourself.
Turn any domain folder of skills into a bounded agentic loop: compile a goal into a verifiable task plan, execute tasks with the domain's own tools, verify every task with machine-run checks, retry with caps, escalate to a human when budgets exhaust, and refuse to close until everything is verified or explicitly waived. Use when you want an agent or subagent to pick up a goal and drive it to a verified close across one of this repo's 20 domains ('run this goal through the engineering harness', 'set up an agentic loop for marketing work', 'make the finance domain self-verifying'). NOT for authoring Claude Code Workflow-tool .js scripts (workflow-builder), N-agent tournaments on one task (agenthub), single-file metric optimization (autoresearch-agent), or discovering published loop recipes (loop-library).
Workbench agent panel system — ef-edit CustomEvent pipeline, registry roll-up, selector grouping, and element property schema. Use when adding new GUI edit capture points, expanding the inspector schema, or continuing development of the EFAgentPanel feature.
A shared, file-based town square where multiple coding agents talk, coordinate, and debate — no server required. Use whenever more than one agent works the same repo (parallel Claude Code or Codex sessions, separate git worktrees, a fleet splitting a task) and they must stay out of each other's way or think together. TRIGGER on phrasings like "coordinate with the other agent/session", "post to / check the agora", "ask the other agents", "leave a message for whoever's working on X", "announce what files you're touching", "is anyone else editing this?", or any time you're about to edit shared code while other agents are live. Also trigger when an agent is stuck and wants a peer's second opinion, or when several agents each drafted a design (an API, a schema, an architecture) and the group needs to compare the proposals and converge on the best one. Works for any agent that can run a Python script, not just Claude Code.
Create an agent platform instance in DMS via Alibaba Cloud OpenAPI. Supports Simple Mode and Advanced Mode. Use this skill when the user wants to provision, deploy, or set up a new Dify instance on Alibaba Cloud DMS.
CLI tools execution specification (gemini/claude/codex/qwen/opencode) with a unified prompt template, mode options, and auto-invoke triggers for code analysis and implementation tasks. Supports configurable CLI endpoints for analysis, write, and review modes.
Query and display in Chinese the number of Codex resets, the remaining time for each reset, expiry times, and reminder times 1 day and 1 hour before expiry. Use when the user asks how many Codex reset credits, reset次数, rate-limit reset credits, credits expiry times, reminder times, or how long each Codex reset credit has before expiration.
Audit and manage the full project context landscape: CLAUDE.md memory hierarchy, project documentation, markdown footprint, and content overlap. Detects project type, scores quality, flags stale docs, and reports total context cost. Trigger with 'audit context', 'audit memory', 'update CLAUDE.md', 'restructure memory', 'session capture', 'check project docs', 'markdown footprint', or 'what docs does this project need'.
Create new agents for the code-forge application. Agents are stored as .md files in the <cwd>/.forge/agents directory with YAML frontmatter (id, title, description, reasoning, tools, user_prompt) and markdown body containing agent instructions. Use when users need to add new agents, modify existing agents, or understand the agent file structure.
Parallelizes work across six patterns — built-in agents, worktree-isolated agents, background agents, custom inline-brief agents, multi-turn agent teams, and parallel task dispatch — choosing teams when agents must share state, subagents when independent. USE WHEN parallel execution, agent team, swarm, spawn agents, fan out, divide and conquer, multi-agent, coordinate agents, custom agents.
Require fresh verification evidence before claiming work is complete, fixed, passing, committed, PR-ready, or ready for handoff. Use before completion claims, commits, pushes, PR creation, issue status changes, delivery summaries, or agent handoffs; do not use for brainstorming or planning that makes no delivery claim.
Comprehensive guide for authoring Claude Code extensions including skills (SKILL.md), hooks, slash commands, and subagents. Use when: creating skills, writing hooks, making slash commands, building subagents, learning Claude Code extensions.
Invoke the locally installed Grok Build CLI from Codex and return Grok's response. Use whenever the user asks Codex to call, consult, run, or delegate a prompt to Grok or Grok Build, including Grok web/X searches and agentic Grok tasks. This Skill covers only the invocation mechanism, process monitoring, and response capture; it does not prescribe the task itself.
Cursor-native offensive security engagement and exploitability autoresearch orchestrator inspired by offensive-claude. Use for an authorized offensive engagement, red-team or pentest workflow, Kill Chain style assessment, scoped web/network/cloud/mobile/AD/bug-bounty offensive plan, or exploitability validation of defensive findings. In validate-findings mode it runs a bounded, user-controlled autoresearch loop over deduplicated findings: several hypothesis passes / cycles of hypothesize, experiment, observe, refine. Use it whenever the user asks to run an autoresearch loop, an iterative or multi-pass hypothesis loop, or to loop or iterate over findings for a set number of cycles. Supports engagement mode and validate-findings mode. Instruction-only: ships no scanners, exploit runners, payload builders, or local scripts.
Run an improve-my-MCP campaign: an autoresearch-style loop that measures the MCP agent experience with the eval harness, picks the highest-impact tool problem from production data, makes one bounded fix, and keeps it only if before/after scores improve. Use when asked to "improve my MCP", run an MCP improvement campaign, fix tool discoverability or descriptions based on evidence, or prepare an eval-backed PR for a tool change. Every shipped change must carry eval evidence; guardrails below are hard rules.
Use this skill when the user says "/li", or when a request relates to content creation but the intent is ambiguous and it is unclear which tool to use. As the main entry point of the li toolbox, it determines intent and routes to the corresponding specialized skill. Should not trigger: when the user's request already clearly matches a specific li-* skill (e.g. "write a script" → li-writer, "refine a topic" → li-topic), trigger that skill directly instead of this entry point. DO NOT trigger for non-content-creation tasks. Use when the user says "/li" or the intent is ambiguous across multiple li-* tools.
Get an independent verdict on an AI-proposed plan, fix, architecture, or claim by verifying its premises against the real artifact, then dispatching a panel of fresh models from different providers that answer independently. Use for "double check", "proof this", "are you sure", "second opinion", "am I right or is he right", "this still isn't working", or "not sure this is the right approach".
Navigates archived ChatGPT or Claude-style conversation exports and a MemFS reference archive on demand. Use when recalling what a past assistant knew, searching old conversations, rendering specific chats, seeding reference memory from export sidecars, or mining historical context without doing a full import.
World-class prompt powerhouse that generates production-ready mega-prompts for any role, industry, and task through intelligent 7-question flow, 69 comprehensive presets across 15 professional domains (technical, business, creative, legal, finance, HR, design, customer, executive, manufacturing, R&D, regulatory, specialized-technical, research, creative-media), multiple output formats (XML/Claude/ChatGPT/Gemini), quality validation gates, and contextual best practices from OpenAI/Anthropic/Google. Supports both core and advanced modes with testing scenarios and prompt variations.
Package a codebase into a compact, structured context bundle for an AI agent. Use when the user wants to "give the whole repo to an LLM", "pack the codebase", hits context limits sharing a project, or asks for a digest/summary of an unfamiliar repo.
Create detailed, phased implementation plans through interactive research and iteration. Use when the user explicitly asks to “create a plan,” “plan the implementation,” or “design an approach” for a feature, refactor, or bug fix. Do not use for quick questions or simple tasks
**WORKFLOW SKILL** — Iteratively improve skill frontmatter compliance using the Ralph loop pattern. WHEN: "run sensei", "sensei help", "improve skill", "fix frontmatter", "skill compliance", "frontmatter audit", "score skill", "check skill tokens". INVOKES: token counting tools, test runners, git commands. FOR SINGLE OPERATIONS: use token CLI directly for counts/checks.
Writer protocol preloaded into the seo-fixer-writer agent — how confirmed SEO/AI-search fixes are applied (git pre-flight, backup, Edit/Write for local diffs, ticketed adapter CLIs for remote targets, re-verify, publish only on a second ticket, rollback) and the findings-array output contract. Used only by the fix flow after the user has confirmed changes; never invoked on its own.
Set up or audit a project for coding agents - the machine checked first with set-up-machine, `AGENTS.md` as the one rules file with `CLAUDE.md` pointing at it, the issue tracker, triage labels, domain docs and release lifecycle, the folder standard's `.gitignore`, and project harness files kept allow-only. Use for a new or existing project, to audit a project's agent setup, or when a skill finds no issue tracker configured.
Installs prerequisite CLI tools (gh, glab, jq), guides GitHub/GitLab authentication, installs uvx, and configures the existing mcp-atlassian MCP server for Jira in OpenCode. Idempotent: safe to rerun the skill. Triggered when the user says “setup,” “install prerequisites,” or “configurer jira.” To build a new MCP server (rather than configure an existing one), use mcp-builder instead.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy
Meta-skill for picking project skills via APM. Invoke ONLY when the user explicitly asks to set up apm.yml, choose which skills a project needs, or evaluate a catalog match — do NOT auto-invoke on routine project-init or apm-management tasks. Two-phase: pick from a curated catalog first; only escalate to broader search/evaluation when the catalog has no fit. Avoids the failure mode of impulse-installing skills you never use or hand-searching GitHub when a vetted answer already exists.
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.