The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill templates the ingestion of untrusted external data through HTTP request headers and body parsing, creating a potential surface for indirect prompt injection.
Ingestion points: The code snippets in SKILL.md demonstrate extracting data from req.headers.authorization and req.json().
Boundary markers: No explicit boundary markers or instructions to disregard instructions embedded in input data are included in the templates.
Capability inventory: The templates demonstrate capabilities including database operations via Supabase, caching via Redis, and logging to the console.
Sanitization: The skill includes a snippet demonstrating Zod validation (z.ZodError) for error handling, though the specific input validation schemas are left to the implementation.

backend-patterns