Agent Skills · Cross-Agent Registry

Q: What is the Agent Skills open standard?

Agent Skills (SKILL.md) is an open format originally released by Anthropic and now adopted by 40+ AI agent platforms via agentskills.io. A skill is a folder containing a SKILL.md file with YAML frontmatter (name, description) plus optional scripts, references, and assets.

Q: How is GuildSkills different from ClaudSkills?

ClaudSkills focuses on the Claude Code ecosystem. GuildSkills covers the broader cross-agent universe. Shared backend, shared Pro subscription, different positioning.

Q: Are these skills authored specifically for each agent?

Most skills work across every compatible client without modification. Skills are listed with full author attribution and a clear note that behavioural compatibility depends on whether the skill uses agent-specific features.

Q: Is GuildSkills affiliated with Anthropic, Nous Research, or Cursor?

No. GuildSkills is an independent community catalog. The SKILL.md format itself is an open standard at agentskills.io.

The compatible clients

Works on every agent that adopts the open SKILL.md standard

The Agent Skills format is open and shared — the same skill file works on Hermes, Cursor, Codex, and the rest of the agent universe. Browse the catalog filtered to any specific client, or sweep the lot.

Why a cross-agent registry

One format. Forty compatible clients. One catalog.

The same SKILL.md file works across every agent client implementing the open standard. The registry, search, ranking, and install layer is the product. The skills stay with their authors.

01 · Catalog

85,000+ skills, one open format

Mined nightly from public GitHub, gists, awesome-lists, and direct submission. Every entry follows the SKILL.md spec from agentskills.io — author attribution, license, and source repo preserved on every page.

02 · Quality Score

Multi-signal ranking, Pro-gated

The free catalog stays unrated — every admitted skill appears equally on browse pages. Pro unlocks a multi-signal Quality Score on every skill based on content structure, anti-trigger discipline, and frontmatter depth. Authors are never paid; ranking cannot be bought.

03 · Install

Per-agent install commands

Each skill page carries the install command for every compatible agent — hermes skills install, cursor add-skill, Codex CLI, Gemini CLI, and the rest. Pro members get one-click install via the desktop app.

04 · Open standard

The spec lives at agentskills.io

GuildSkills is the registry; agentskills.io is the format spec (originally released by Anthropic, now adopted by 40+ AI agent platforms). We catalog and rank — the format itself is open and governed by the community.

How it works

From open spec to running agent — three steps

The same workflow regardless of which agent you're using. The SKILL.md format handles the rest.

Step 01

Browse or filter

Filter by agent client, category, tag, or use case. Pro members see the Quality Score on every skill.

Step 02

Copy the install command

Each skill page carries the install command for every compatible agent. One line, copy-paste, done.

Step 03

Run it in your agent

The agent loads the skill via progressive disclosure when a matching task fires. Author attribution stays with the skill.

Step 04

Unlock Pro Quality Score

One Pro subscription works on GuildSkills and ClaudSkills. Multi-signal ranking, one-click install via desktop app, curated theme packs.

Frequently asked

Questions worth asking

What is the open standard, how does cross-agent compatibility work, and why a separate brand from ClaudSkills.

What is the Agent Skills open standard?

Agent Skills (SKILL.md) is an open format originally released by Anthropic and now adopted by 40+ AI agent platforms via agentskills.io. A skill is a folder containing a SKILL.md file with YAML frontmatter (name, description) plus optional scripts, references, and assets. Agents load skills via progressive disclosure: discovery (name + description at startup), activation (full instructions when task matches), and execution (optionally run bundled scripts).

Which agents are compatible with skills in GuildSkills?

Skills following the SKILL.md format work without modification across Hermes Agent, Cursor, OpenAI Codex, Gemini CLI, OpenCode, OpenHands, Goose, Letta, Roo Code, GitHub Copilot, VS Code, Junie (JetBrains), Tabnine, Firebender, Amp, Factory, Kiro, Snowflake Cortex Code, Databricks Genie Code, Spring AI, Mistral Vibe, Workshop, Trae (ByteDance), Emdash, Qodo, Laravel Boost, OpenClaw, Piebald, Ona, VT Code, Command Code, Agentman, Vita, Superconductor, pi, nanobot, fast-agent, bub, Google AI Edge Gallery, Mux, and Claude Code.

How is GuildSkills different from ClaudSkills?

ClaudSkills focuses on the Claude Code ecosystem specifically. GuildSkills covers the broader open Agent Skills standard across 40+ compatible clients. The two share a backend — same catalog, same Quality Score, same Pro subscription — but present them under different positioning. Pro is a single purchase that unlocks scoring on both sites.

How do I install a skill on a specific agent?

Each per-agent landing page carries client-specific install commands. Quick reference: Hermes Agent uses hermes skills install <repo>; Claude Code uses ~/.claude/skills/ directory; Cursor uses .cursor/skills/; Codex CLI uses codex skill add. See the per-agent pages for full details.

Are these skills authored specifically for each agent?

The SKILL.md format is an open standard, so most skills work across every compatible client without modification. Skills are listed with full author attribution preserved and a clear note that behavioural compatibility depends on whether the skill uses any agent-specific features. Authors can request opt-out by email at any time.

Is GuildSkills affiliated with Anthropic, Nous Research, or any agent vendor?

No. GuildSkills is an independent community catalog. Anthropic released the SKILL.md format as an open standard at agentskills.io — that's the spec home. GuildSkills is the cross-agent registry built on top of the spec, independent from every agent vendor.