Our take
- Activity
- 53
- Community
- 75
- Issues
- 21
- Pull requests
- 59
+8,242 stars in the last 30 days
- Merges more pull requests than 98% of the projects we track
- More contributors than 77% of the projects we track
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
167,801Forks
27,221Change since Nov 11, 2026: +121 stars·+28 forks
Overview
agency-agents is a repository of Markdown instruction files for AI agents spanning software engineering, design, marketing, security, finance, game development, and other fields. Each file defines a specialist's role, working approach, and expected deliverables, so users can select focused instructions for a particular task instead of starting with a generic assistant prompt.
The repository also includes shell scripts to convert and install selected agents into supported AI coding tools. It is suited to people who already use those tools and want a broad, editable roster of role-specific instructions they can browse, adapt, or install.
Key Features
- Organizes specialist agent instructions into professional divisions, including engineering, design, product, security, marketing, and support.
- Provides agent files as Markdown that can be read, copied, or adapted directly.
- Includes scripts to convert agent files and install them into supported tools.
- Supports selecting agents by tool, division, or individual agent through installer options.
- Offers an interactive installer and a dry-run option for checking selections.
- Documents a desktop app as an alternative way to browse and install the roster.
Use Cases
- A developer wants a focused frontend or database specialist in a compatible coding assistant, and installs only that agent.
- A small team wants reusable role instructions for recurring work such as code review, testing, or technical writing.
- A product or marketing practitioner wants to adapt a domain-specific workflow without building a prompt from scratch.
- A team evaluating AI workflows wants to browse and compare a wide range of specialist roles before choosing what to adopt.
Getting Started
For Claude Code, install the agents with:
./scripts/install.sh --tool claude-code
See the README for other tools, selection options, and manual setup.
Alternatives
- awesome-agent-skills: It curates task-focused skills from many sources and supports several coding assistants, rather than centering on professional-role instruction files.
- agent-plugins: It packages skills, tool connections, and specialist roles for specific coding assistants, rather than primarily distributing reusable role instruction files.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| agency-agents | Shell | MIT | 156k | Not checked yet |
| awesome-agent-skills | not specified | MIT | 35.2k | Active |
| agent-plugins | Shell | Apache-2.0 | 2 | New |
Considerations
- These files provide instructions for AI tools, not standalone agents or models. Their results depend on the host tool and model interpreting the instructions.
- The roster is broad, so selecting and maintaining a useful subset may take more effort than using a small, purpose-built prompt collection.
- Installation and conversion use shell scripts. Review scripts and generated files before using them in environments with specific security or policy requirements.
- The README notes an OpenCode runtime limit that can silently drop agents when too many are installed. Use a subset for that tool and check its current upstream behavior.
Found this useful?
Share it with someone who would like agency-agents.