The open registry of production-ready skills, agents, plugins, and contexts for AI coding assistants.
Drop-in components built and refined by the community. No bloat. No boilerplate.
| Type | What it does |
|---|---|
| Skills | Behavioral instructions loaded into Claude on demand |
| Agents | Autonomous subagents for multi-step tasks |
| Plugins | Integrations with external tools and services |
| Contexts | Background knowledge files (reference docs, domain guides) |
Install the whole registry (recommended):
/plugin marketplace add msdakot/ai-foundaryManual clone:
git clone https://github.com/msdakot/ai-foundary.git ~/.claude/plugins/ai-foundaryOne-liner:
curl -fsSL https://raw.githubusercontent.com/msdakot/ai-foundary/main/scripts/install.sh | bashInstall a single component via npx:
npx ai-foundry add <type> <name># Install a skill
npx ai-foundry add skill <listed-skill>
# List everything available
npx ai-foundry listThis copies the component into your project's .claude/ directory and registers it in settings.json.
| Name | Description |
|---|---|
| codeassist-guardrails | Behavioral guardrails for LLM coding — prevents silent assumptions, overengineering, and collateral damage |
| skill-observer | Background observer that watches conversation patterns, detects repeated corrections, and suggests new skills or CLAUDE.md rules for human approval |
| content-curator-research | Autonomous 3x/day AI intel sweep across HackerNews, GitHub Trending, ArXiv, Google AI Blog, and X — appends structured daily reports to a GitHub repo |
| content-curator | Drafts LinkedIn posts and newsletter sections grounded in real signals from the latest intel report, with hook library and reverse-engineer mode |
| Agent | Description | Sub-agents |
|---|---|---|
| ai-data-agents | Specialist sub-agents for end-to-end AI/ML work spanning research, data, modeling, and deployment | deepresearch, autoresearch, ai-scientist, data-scientist, ml-engineer, data-engineer, mlops-engineer, llm-architect, prompt-optimization, nlp-engineer, computer-vision-engineer, recommendation-engine, vector-database-engineer, data-visualization |
| devops-agent | Specialist sub-agents for the full software development lifecycle — from idea to deployed, reviewed code | feature-brainstorm, spec-writer, architect, diagram-definition, task-planner, developer, language-expert, code-reviewer, mcp-developer, devops-infra |
None yet — contribute one
None yet — contribute one
-
Copy the component folder into your project:
.claude/ └── .agents/ └── skills/ └── <skill-name>/ └── SKILL.md -
Register it in
.claude/settings.json:{ "skills": [ ".claude/.agents/skills/<skill-name>" ] } -
Skills with hooks (e.g.
skill-observer) also require copying the hook scripts to~/.claude/hooks/and registering them in~/.claude/settings.json:cp .claude/.agents/skills/<skill-name>/hooks/*.sh ~/.claude/hooks/ chmod +x ~/.claude/hooks/*.sh
Then add to
~/.claude/settings.json:{ "hooks": { "Stop": [{ "hooks": [{ "type": "command", "command": "bash ~/.claude/hooks/<skill-name>-stop.sh", "timeout": 10 }] }] } }The
npx ai-foundry addcommand handles all of this automatically.
See CONTRIBUTING.md.
