Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Set up and improve Harness engineering for AI agent-friendly codebases (including AGENTS.md, docs/, lint rules, eval systems, project-level prompt engineering). Trigger scenarios: setting up a new/empty project for an AI agent, creating AGENTS.md or CLAUDE.md, questions about Harness engineering, making an agent work more efficiently on a codebase. Also triggers when the user is frustrated or complains about agent quality (e.g. 'the agent always ignores the spec', 'it never follows instructions', 'why does it always get X wrong', 'the agent is broken') — because poor agent output almost always means missing Harness, not a model problem. Covers: context engineering, architectural constraints, multi-agent collaboration, evaluation, long-running task Harness, the agent loop (autonomous operation, verifiers, stop conditions), agent memory (cross-session memory, memory stewardship), and diagnosing agent quality issues.
Harness Engineering phase 1: scan an existing project and generate AGENTS.md (a directory file) and a complete docs/ knowledge base structure. Use this skill immediately when the user wants to add agent support to a project, help AI understand my project better, start harness engineering, create AGENTS.md, set up an agent documentation structure, or make Claude Code work better. Also applies when the user says help me organize the project docs for agent use, I want to start developing with AI agents, sort out what business problems this project solves, or asks to create business-solution.md.
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).
Create custom slash commands for Claude Code with markdown and YAML frontmatter. Use when building workflow automations, creating reusable prompts, or defining custom /commands. Use for slash commands, argument handling, file references, bash execution.
Create event hooks for Claude Code that trigger on tool calls, prompts, and lifecycle events. Use when automating workflows, blocking dangerous commands, protecting files, or adding notifications. Use for hooks, PreToolUse, PostToolUse, event handling, automation.
Goal-driven conductor for the whole idea-to-published-paper lifecycle. Takes a research idea plus experiments and a target ("submit to NeurIPS 2026", or "help me pick a venue") and runs a goal->plan->execute->verify->reflect loop. It plans the stages, invokes the right lifecycle skills in order (literature-review, write-abstract, preflight-check, simulate-reviewers, and the rest), checkpoints with the author before each stage, and re-verifies the live CFP via parse-cfp rather than trusting a cached profile. Use when a researcher says "take my idea to a submitted paper", "run the whole pipeline", "orchestrate the paper for venue X", "what's the plan from here", or "what's done and what's next". Copilot, not autonomous author. It stops to ask at every stage gate, never fabricates results or citations, and never submits. Trigger words - orchestrate, pipeline, end to end, whole paper, idea to submission.
Structured planning workflow for any coding task. Use at the start of every new feature, bug fix, refactor, or implementation request. Analyzes the project, asks up to 5 clarifying questions, creates a TODO.md, gets user approval, then executes task by task. Never writes code before a plan is approved.
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
Maintains project context and progress tracking across Claude sessions. Use at session start to load context, on session end to save progress. Triggers: load project context, save context, end session, what was I working on, switch to project, done for today.
The reflection agent. Use it right after another skill produces an artifact such as a rewrite from polish-prose, a rebuttal, an abstract, a preflight fix, or a tailored draft to critique that output against the stated goal, the paper's .paper-memory/profile.yml, the relevant rubric, and past .paper-memory/lessons.md, then decide whether it actually improved. Trigger words include reflect, self-critique, did this get better, did my edit help, regression check, review my own output, iterate, refine, second pass, sanity-check the rewrite, and "is the new version actually better". Runs a bounded self-refine / Reflexion loop, guards against regressions, and appends durable, deduplicated, dated lessons to .paper-memory/lessons.md via a stdlib script so the toolkit gets sharper each paper. Reflection adds cost; it only pays off when there is a measurable target.
Initialize the repository’s writing standards and engineering skills: add writing requirements and coding guidelines to AGENTS.md, and configure the issue tracker, triage labels, and domain documentation. Run manually when setting up a repository for the first time.
Evaluate and benchmark Salesforce skill quality. Compares AI-generated code with vs without skill context, scores against a Salesforce-specific rubric (security, governor limits, bulkification, patterns, completeness), and produces a comparison report. Use to run benchmarks, verify skill value, or check Apex code quality. Activate on mentions of "evaluate skills", "benchmark", "skill quality", "run eval", or "compare with/without skills".
Pre-flight checks, reliability acceptance testing, and format review tooling for Skill creation. Use this skill when a user creates, substantially reworks, or reviews a Skill and needs to identify unstable instruction following, output drift, validation-modality mismatches, or missed constraints in older Skills, or to check Harness contracts, candidate binding evidence, fault injection, directory structure, business flows, and security risks. Do not use for: replacing domain validators, code review, application functional testing, or general programming tasks.
Automated retrospectives — captures learnings at EPIC completion and on manual invocation. EPIC-scoped retros embed a Retrospective section in the EPIC artifact. Cross-epic and time-based retros produce standalone retro docs. Triggers on: 'retro', 'retrospective', 'post-mortem', 'lessons learned', 'debrief', 'what worked', 'what didn't work', 'what did we learn', 'reflect', or automatically after EPIC terminal transitions.
Turn the prompt supplied with this skill into a concise, auditable Codex Goal or explain why a Goal is not the right fit. Use when the user asks to draft, formulate, rewrite, tighten, or create a `/goal` from a plain-language task, especially for multi-step work that needs a durable objective, evidence-based completion, constraints, iteration policy, and a default adversarial review loop.
One-click deployment of camoufox anti-detection browser with modified agent-browser. Patches agent-browser to auto-detect camoufox/firefox from executable path instead of defaulting to chromium. Includes SkillsI integration for seamless browser automation workflows.
Defines behavioral and cognitive standards for senior AI engineering agents. Use when configuring agent reasoning protocols, memory management, or context engineering strategies. Use for autonomous reasoning, tiered memory systems, verifiable goal execution, multi-agent orchestration, and token optimization.
Orchestrate multi-agent teams with defined roles, task lifecycles, handoff protocols, and review workflows. Use when: (1) Setting up a team of 2+ agents with different specializations, (2) Defining task routing and lifecycle (inbox → spec → build → review → done), (3) Creating handoff protocols between agents, (4) Establishing review and quality gates, (5) Managing async communication and artifact sharing between agents.
Generate, validate, and explain `auth.md` files — the open protocol for AI agent registration. Use when making apps agent-ready, generating Protected Resource Metadata (RFC 9728), validating auth.md files, or implementing agent registration endpoints. Triggers on "auth.md", "agent registration", "agent auth", "make my app agent-ready", "ID-JAG", "identity_assertion", "service_auth", "protected resource metadata", "agentic registration".
Use when running Ralph-style iterative autonomous development. Triggers on /ralph or /loop commands, when autonomous iterative development is needed, when a project has specs and an implementation plan ready for iterative execution, or when deterministic context loading with subagent delegation and dual-condition exit gates is required. Orchestrates PLANNING, BUILDING, and STATUS cycles.
Analyze coding sessions to detect corrections and preferences, then propose targeted improvements to Skills used in the session. Use this skill when the user asks to "learn from this session", "update skills", or "remember this pattern". Extracts durable preferences and codifies them into the appropriate skill files.
Run isolated eval and grading calls using CC 2.1.81 --bare mode. Constructs claude -p --bare invocations for skill evaluation, trigger testing, and LLM grading without plugin/hook interference. Use when running eval pipelines, grading skill outputs, benchmarking prompt quality, or testing trigger accuracy in isolation.
Build a Loop for PostHog Desktop: a workflow that creates an AI task each time its trigger fires, optionally followed by a Slack or email notification with the task's result. Use when asked to create, set up, or change a loop, a recurring agent, a scheduled task, or an automation that runs an AI task when a GitHub, Slack, or PostHog event happens. Covers the exact graph, trigger configs, schedule presets, the notify step, structured output, and the test-run steps.
Expert in making multi-agent systems resilient. Specializes in detecting loops, hallucinations, and failures, and implementing self-healing workflows. Use when designing error handling for agent systems, implementing retry strategies, or building resilient AI workflows.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Orchestrate coding work by delegating well-specified implementation tasks to xAI's Grok Build CLI (grok) running headlessly, while the coding assistant plans, writes the task specs, reviews every diff, and owns the result. Use when user says: 'use grok', 'grok build', 'delegate to grok', 'have grok implement', 'have grok execute', 'have grok build', 'send to grok', 'execute this plan with grok'. Executes a Markdown implementation plan task-by-task, or ad-hoc tasks with an inline spec.
Use when the user wants to build or change something but the request is vague or open-ended: “help me figure out what to build,” “I'm not sure what it should do,” “add X” with no details, or no clear definition of done. Runs a short Socratic interview to turn the idea into goals, constraints, and acceptance criteria before any code is written. Not for simple questions, small clear edits, or an existing Seed.
Audit a repo against Cloudflare/isitagentready signals for AI agent readiness: robots.txt, llms.txt, Markdown negotiation, MCP/API catalogs, OAuth discovery, WebMCP, and browser-agent usability. Includes explicitly scoped readiness audits; not generic SEO or standalone standards documentation questions.
Cost-optimize AI agent operations by routing tasks to appropriate models based on complexity. Use this skill when: (1) deciding which model to use for a task, (2) spawning sub-agents, (3) considering cost efficiency, (4) the current model feels like overkill for the task. Triggers: "model routing", "cost optimization", "which model", "too expensive", "spawn agent".
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.