# ai_stp

> The AI setup registry for skills, MCP, hooks, and subagents — not another skills.sh. The web is the catalog; the agent installs through the CLI.

## Primary resources
- [Catalog](https://ai-stp.aiguild.space/en/catalog): public components and setups
- [Human documentation](https://docs.nddev.asia): product, CLI, catalog, trust and troubleshooting guides
- [OpenAPI](https://ai-stp.aiguild.space/openapi.json): HTTP contract served by the deployed API
- [Provider-info schema](https://ai-stp.aiguild.space/schemas/provider-protocol/v3/provider-info.json): JSON Schema for protocol v3 provider-info
- [Agent instructions](https://ai-stp.aiguild.space/agents.md): safe machine onboarding
- [Extended context](https://ai-stp.aiguild.space/llms-full.txt): terminology, trust model and supported harnesses
- [Catalog manifest](https://ai-stp.aiguild.space/llms/catalog.ndjson): paginated active subjects
- [Content](https://ai-stp.aiguild.space/en/content): articles, product news, changelog and release notes


## Rules
- Public catalog metadata is readable without an account.
- author_verified is not proof that component content is safe.
- Experimental objects require explicit request-scoped consent and are not auto-installed.
- Never infer private object existence from public responses.
- Use stable ids and exact versions; version numbers may be non-contiguous.