Explore AI Agent Skills & Claude Prompts
Discover open-source agent skills for Claude Code, Codex, ChatGPT, and any tool that uses SKILL.md.
Enter through keywords, occupations, creators, and GitHub sources to see what kinds of skills are emerging across domains.
Use the same catalog through the API
Connect 381,784 public skills to your own search, analytics, or agent workflow with the REST API.
Querying local SQLite index...
cj-portability-audit
by jcl2018Static dependency lint for declared skill portability. Compares each catalog skill's declared `portability` field against its ACTUAL executed repo-local dependencies (root scripts/*.sh helpers, root config, CLAUDE.md, the manifest `.source` reach-back) using a strict tier ladder (standalone < local-only < workbench), an EXECUTED-vs-documented precision rule, bundled-own-script + scoped self-resolution-preamble carve-outs, and an optional `portability_requires` accepted-deps field. Emits a per-skill verdict (portable / portable-with-notes / findings:<list>). Engine-in-script; also wired into validate.sh as an advisory check (exit 0 in v1; PORTABILITY_STRICT=1 flips to hard-fail). Workbench-only. Use when: 'audit skill portability', 'check declared-vs-actual dependencies', 'is this skill really standalone'.
cj-qa-work-item
by jcl2018QA a CJ_personal-workflow work-item against its test rows. User-stories get smoke tests + a fresh-context E2E subagent per TEST-SPEC; defects and tasks run their test-plan rows as smoke-equivalent. Every green path then runs the Step 8.6 audit block — refresh spec/test-spec-custom.md (8.6a) + spec/doc-spec-custom.md (8.6b) ALWAYS inline, then the three-stage audits (8.6c /CJ_doc_audit + 8.6d /CJ_test_audit) inline UNLESS the dispatch prompt carries the literal DEFER_AUDIT: true directive. Standalone runs (no directive): audits run inline and findings ride the GREEN RESULT's AUDITS=doc:..,test:.. field + a fenced AUDIT_FINDINGS block (the cj_goal post-QA checkpoint owns the Continue/Halt decision; findings never flip QA red). Orchestrator-driven runs with DEFER_AUDIT: true: 8.6c/8.6d are SKIPPED, the RESULT reads AUDITS=deferred with no AUDIT_FINDINGS block, and the orchestrator runs the audit at the post-sync point. Writes findings to tracker journal, transitions Phase 2 QA-owned gates, refuses on incomplete
cj-scaffold-work-item
by jcl2018Scaffold a CJ_personal-workflow work item from an /office-hours design doc. Reads design + templates + manifest + WORKFLOW.md, produces a compliant work-item directory tree with all required artifacts. Runs /CJ_personal-workflow check at boundaries; idempotent (re-run on same input is NO-OP).
cj-suggest
by jcl2018Print a ranked top-5 of next-up work items from TODOS.md and tracker frontmatter. Internal phase-step skill rows (CJ_scaffold-work-item, CJ_implement-from-spec, CJ_qa-work-item, *-workflow validators) are filtered by default; pass --include-internal to surface them. Optional --for-skill / --limit flags pre-filter and extend the candidate window for downstream callers like /CJ_goal_todo_fix.
cj-system-health
by jcl2018~/.claude/ health dashboard with dependency graph and usage trends. Scans installed skills, builds a dependency graph, checks filesystem health, surfaces skill usage analytics with behavioral topology overlay, and optionally invokes waza for config hygiene. Produces a scored report with trend tracking.
cj-goal-defect
by jcl2018Bug-description-to-shipped-fix orchestrator (F000027 `defect` verb; experimental). Takes a plain bug description with NO pre-existing defect dir, scaffolds a throwaway `.inbox/<slug>/DRAFT.md`, root-causes it via /investigate as an Agent subagent (sentinel-wrapped JSON, Iron-Law: no root cause ⇒ HALT, nothing promoted or shipped), then on a populated root cause writes RCA + test-plan, promotes the draft to a canonical `work-items/defects/uncategorized/D000NNN_<slug>/` dir (D-ID minted only after the Iron-Law gate passes), runs /CJ_qa-work-item (leaf subagent; with DEFER_AUDIT: true so the three-stage audit is deferred to the post-sync point), makes an idempotent pre-doc-sync commit, folds doc updates via /CJ_document-release (Step 5.5 doc-sync; halt-on-red), runs ONE combined read-only post-sync doc/test audit (Step 5.6), surfaces the post-QA audit checkpoint AUQ on that POST-sync report (Step 8.5 — ALWAYS; Continue past findings journals [qa-audit-waived], Halt journals [qa-audit-declined] / halted_at_qa_aud
cj-document-release
by jcl2018Workbench wrapper around upstream /document-release. Reads the MERGED two-tier doc-spec registry (the general spec/doc-spec.md, resolved spec/-then-root, + the optional spec/doc-spec-custom.md overlay; self-bootstraps a missing general file from the portable seed INTO spec/doc-spec.md; stub-scaffolds any missing declared doc — spec/test-spec.md special-cased via test-spec.sh --seed), adds a --docs <comma-list> subset flag for per-invocation doc filtering (best-effort, documentation-only), a halt-on-red contract that emits [doc-sync-red] on upstream failure, and an auto-commit step gated by a doc-only whitelist DERIVED from the merged registry (non-whitelist writes HALT with [doc-sync-non-doc-write]). A missing/invalid registry HALTs with [doc-sync-no-config] BEFORE any audit. Invoked inline by the cj_goal orchestrators at Step 5.5 — between the QA pass + post-QA audit checkpoint and /ship — so doc updates fold into the same code PR rather than chasing them post-merge.
cj-repo-init
by jcl2018Detect which CJ_ skills are deployed, verify each one's per-repo prerequisites (cj-document-release.json, CJ-DOC-RELEASE.md, TODOS.md, work-items/ tree), print a health table, and on one confirm scaffold the missing repo-level prerequisites from generic portable seeds. Standalone utility — in-place, no worktree/ship. Use when: 'set up this repo for the CJ skills', 'init repo prerequisites', 'make this repo ready for CJ_', 'bootstrap repo config', 'verify repo prerequisites'.
cj-goal-task
by jcl2018Small-ad-hoc-task-to-reviewable-PR orchestrator (F000054 `task` verb; experimental). The lightweight sibling of /CJ_goal_feature: takes a plain free-text `"<small task>"` (refine a doc, add a file, clean up some files, a one-line fix), runs a HARD complexity gate (refuses design-rework topics → routes to /CJ_goal_feature, refuses bug/investigation topics → routes to /CJ_goal_defect, refuses explicit-large-scope topics; HALTs as halted_at_too_complex), creates a `cj-task-*` worktree, then SILENTLY (one checkpoint AUQ — the QA audit findings) bash-scaffolds a `type: task` work-item (T-ID) directly from the topic via scripts/cj-task-scaffold.sh — NO /office-hours, NO design doc, NO pre-existing TODOS row — and dispatches /CJ_implement-from-spec → /CJ_qa-work-item (with DEFER_AUDIT: true — the three-stage audit is deferred to the post-sync point) as depth-≤2 leaf Agent subagents, makes an idempotent pre-doc-sync commit, folds doc updates via /CJ_document-release INLINE (Step 5.5 doc-sync), runs ONE combined read-
cj-goal-feature
by jcl2018One-line-topic-to-reviewable-PR feature orchestrator (F000027 `feature` verb; experimental). Takes a plain feature topic, creates a `cj-feat-*` worktree, runs /office-hours INLINE (the one interactive design phase; emits an APPROVED design doc — on not-APPROVED/abandoned it HALTs), then shows a design-summary approval gate in chat (a concise digest of the APPROVED doc + a single go/no-go before the autonomous build budget is spent — Abort HALTs as halted_at_design_gate and preserves the doc for resume), then SILENTLY (one checkpoint AUQ — the QA audit findings — past the gate) dispatches /CJ_scaffold-work-item → /CJ_implement-from-spec → /CJ_qa-work-item (with DEFER_AUDIT: true — the three-stage audit is deferred to the post-sync point) as depth-≤2 leaf Agent subagents, makes an idempotent pre-doc-sync commit, folds doc updates via /CJ_document-release INLINE (Step 5.5 doc-sync), runs ONE combined read-only post-sync doc/test audit (Step 5.6), surfaces the post-QA audit checkpoint AUQ on that POST-sync report
cj-goal-todo-fix
by jcl2018Drain TODOs from TODOS.md into shipped PRs. Default mode (no args) drains up to 10 easy-fix TODOs end-to-end via /CJ_implement-from-spec + /CJ_qa-work-item + /ship + /land-and-deploy. Pass a T-ID or fragment for single-TODO mode; --max-drain N caps, --dry-run previews, --quiet for cron / /schedule consumers. /ship Gate #2 still fires per drained PR (autonomy ceiling). Use when: 'fix this TODO', 'clear the TODO backlog', 'auto-resolve TODOs', 'drain TODOs'.
cj-implement-from-spec
by jcl2018Implement a CJ_personal-workflow work-item from its input artifacts. Reads per-type spec (SPEC+DESIGN for user-stories, RCA+test-plan for defects, TRACKER+test-plan for tasks; features delegate to a child user-story via AUQ), plans against Components Affected / Data Flow, writes code via Read/Edit/Write. Sensitive-surface AUQ for catalog/manifest/validator edits; propose-and-confirm by default (--auto for trivial ≤2-file changes). Idempotent. Use when: 'implement this work-item', 'write the code for this spec'.
Browse Agent Skills by Occupation
23 major groups · 867 SOC occupations
Browse by Category
Explore agent skills organized by their primary use case
Explore the agent skills ecosystem by occupation and creator
SkillMD is not just a keyword search box. It is an open map that organizes public skills by occupation, creator, and repository, helping you see which workflows, judgment criteria, and domain habits people are writing for AI agents.
Then follow creators and GitHub repositories back to the source: compare the skills a team maintains, whether the repo is active, and how the README frames the work before you open, install, or reuse anything.
Use it three ways: learn an unfamiliar field by occupation, study how creators organize skills, then use source context to decide what is worth opening or reusing.
01 Map a field
Browse 23 occupation groups and 867 SOC roles to learn what skills exist in adjacent domains and how they break down real work.
02 Follow creators
Use creator and repository pages to inspect maintained skill collections, recent updates, and source context before trusting a result.
03 Search with sources
Search 1.7M+ collected skills, then use occupation tags, creators, and GitHub source context to decide what is worth opening.
Start with the occupation map, then follow creators and repositories back to real code. SkillMD helps explain why a skill is worth opening, not only what it is named.
Standardizing Agent Capabilities with SKILL.md and Model Context Protocol (MCP)
In the rapidly evolving landscape of artificial intelligence, LLM agents (Large Language Model agents) have transitioned from simple text predictors to autonomous problem solvers. To orchestrate complex, multi-step agentic workflows, developers require a standardized format to specify agent capabilities, prompt instructions, system rules, and database bindings. This is where SKILL.md and the Model Context Protocol (MCP) have emerged as standard developer paradigms. SkillMD serves as the central directory for indexing, exploring, and sharing these critical agent configurations.
Our open-source registry currently tracks over 1.7 million collected SKILL.md configurations and system prompts. By compiling agent configurations from active developers on GitHub, we bridge the gap between prompt engineering research and production execution. Whether you are building agents with Anthropic's Claude Code, OpenAI's GPT-4, Google's Gemini, or local models using Ollama and LlamaIndex, standardized skill definitions ensure your agents behave predictably across different runtime environments.
What is the Model Context Protocol (MCP)?
The Model Context Protocol (MCP) is an open-source standard designed to connect LLMs to data sources, developer tools, and external environments. MCP establishes a bidirectional communication channel between client applications (like Cursor, Claude Desktop, or custom agent systems) and servers hosting data or capabilities. Standardizing instructions via SKILL.md enables LLMs to query databases, read local files, execute terminal commands, and integrate third-party APIs. SkillMD allows you to find ready-to-run MCP servers and prompt instructions for various occupations and technical tasks.
The Structure of a Professional SKILL.md File
A valid SKILL.md configuration is designed to be easily read by humans and parsed by LLMs. It contains precise system instructions, trigger conditions, required parameters, and execution examples. Below is the typical architectural blueprint of a professional agent skill:
- Metadata & Core Scope: Declares the name of the skill, author details, target models, and a description of the capability.
- Triggers & Intent Detection: Details semantic triggers that help the agent decide when to invoke this skill.
- System Prompts: Explicit system-level instructions that direct the agent's behavior, personality, safety guardrails, and formatting preferences.
- Capabilities & Tools: Lists the files, databases, or APIs the agent must access to complete the tasks.
- Few-Shot Examples: Demonstrates real inputs and outputs, helping the model generalize behavior through in-context learning.
Optimizing Agent Workflows for Modern LLMs
Writing effective agent skills requires deep knowledge of prompt engineering. With the release of advanced reasoning models like Claude 3.5 Sonnet, ChatGPT o1, and DeepSeek-V3, prompt templates must focus on structured thinking. Developers are encouraged to use XML tags (e.g., <thought>, <context>, and <rules>) to isolate execution boundaries. Standardized prompts prevent agents from suffering from context drift, ensuring that long-running tasks remain aligned with the initial system parameters.
Exploring by SOC Occupations and Creator Profiles
What makes SkillMD unique is its taxonomy. Instead of simple text search, we parse and organize files according to the Standard Occupational Classification (SOC) system. This means you can discover skills written for Computer and Mathematical roles, Business and Financial operations, Legal, Design, and and Educational Instruction fields. By tracking creator profiles, developers can study how different teams organize their custom instructions, compare version updates, and fork public configs for specialized enterprise use cases.
SkillMD operates as a high-performance index running on a fast Go backend and a highly responsive Astro SSR frontend. All search queries execute in milliseconds, featuring smart debouncing to prevent multiple API requests while keeping user data secure. Join our community of developers to standardize your AI agent instructions and optimize your LLM prompting workflows today.
Frequently Asked Questions
A practical guide to agent skills: what they are, how to inspect them, and how SkillMD helps you explore the ecosystem.