The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing 'litestar-mcp' via pip. This package is the official tool provided by the vendor 'litestar-org' (the author of the skill) and is hosted in their verified GitHub repository. This is standard and safe installation behavior for this ecosystem.- [DATA_EXFILTRATION]: While the skill's purpose is to expose API data to external AI agents, it provides comprehensive security mechanisms to control this flow. It supports OAuth 2.1 with PKCE, Litestar guards for access control, and explicit allowlists ('include_tags') to ensure only intended data is shared.- [PROMPT_INJECTION]: The skill identifies a potential attack surface for indirect prompt injection via tool outputs from the linked API. However, it mitigates this by providing guidance on robust input validation using Litestar DTOs and minimizing the exposed surface area. Evidence Chain: 1. Ingestion points: Data returned from user-defined Litestar route handlers (SKILL.md). 2. Boundary markers: Structured JSON-RPC 2.0 response format. 3. Capability inventory: Execution of any route handler exposed as a tool (SKILL.md). 4. Sanitization: Use of Pydantic/DTO validation and recommendation of explicit route allowlists.

litestar-mcp