agent-skills

agent-skills

A collection of reusable agent Skills — portable, repeatable workflows that any Claude Code (or compatible) session can discover and run.

Inspired by the structure of addyosmani/agent-skills.

Licensing

This project is dual-licensed:

• Community use is available under the GNU Affero General Public License v3.0 — see LICENSE.
• Commercial, proprietary, closed-source, SaaS, hosted, or internal business use requires a separate written commercial agreement — see COMMERCIAL_LICENSE.md.

Use without a signed commercial agreement is governed solely by the AGPLv3, including its source-availability and network-use obligations. To obtain a commercial license, contact [email protected].

Install as a Claude Code plugin

This repo is also a Claude Code plugin and its own single-plugin marketplace.

/plugin marketplace add https://github.com/acumen-ai-org/agent-skills
/plugin install agent-skills@acumen-agent-skills

The skill is then available as /agent-skills:content-to-image. To develop locally without installing: claude --plugin-dir .. Schemas, versioning, and the reason the marketplace is named acumen-agent-skills (the name agent-skills is reserved by Anthropic) are in docs/plugin-and-marketplace.md.

Quick start

1. Point your agent at this repo (or symlink skills/ into your project).
2. The agent loads each Skill's name + description at startup.
3. When a request matches, the agent reads that Skill's SKILL.md and follows it — pulling in references/ and scripts/ only as needed.

Layout

agent-skills/
├── .claude-plugin/           # plugin.json + marketplace.json (installable plugin)
├── CLAUDE.md                 # entry point for agents working in this repo
├── README.md                 # this file
├── LICENSE                   # AGPLv3 (community use)
├── COMMERCIAL_LICENSE.md     # commercial/proprietary use terms
├── NOTICE                    # copyright + dual-license notice
├── .gitignore                # ignores runtime state + build/OS noise
├── docs/                     # how to author + how the repo is structured
├── skills/
│   └── content-to-image/     # a Skill: SKILL.md + references/ + scripts/
├── agents/                   # roles shared across more than one Skill
├── references/               # repo-level shared references  (empty until needed)
├── hooks/                    # session/agent lifecycle hooks (empty until needed)
└── scripts/                  # repo-level maintenance scripts (empty until needed)

references/, hooks/, and scripts/ are conventional repo-level locations. They start empty, so git does not track them until they hold a file — create each on first use. See docs/repository-structure.md.

Authoring

Start at docs/README.md. The non-negotiables:

• A Skill is self-contained. No Skill depends on a privately-registered subagent; per-Skill roles live in skills/<name>/references/.
• SKILL.md is concise (< 500 lines) and uses progressive disclosure.
• Deterministic, fragile steps are scripts/, not prose.

See docs/skills-vs-agents.md for when to reach for a Skill vs a subagent, and docs/authoring-skills.md for the full checklist.