The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing untrusted data from HTTP requests (headers, JSON payloads, and query parameters). While it includes validation examples like Zod, improper implementation of these patterns in an AI-driven backend could lead to indirect prompt injection if user-provided content is interpolated into agent prompts without sanitization.
Ingestion points: The patterns demonstrate reading from req.headers.authorization, request.json(), and query filters.
Capability inventory: Patterns include database operations via Supabase and external API calls via fetchFromAPI().
Sanitization: The errorHandler snippet includes handling for z.ZodError, encouraging input validation.
Boundary markers: Not applicable for code snippets, though best practices for validation are demonstrated.
[SAFE]: The skill follows industry best practices for secret management by referencing environment variables for sensitive data, such as process.env.JWT_SECRET, rather than hardcoding credentials.

backend-patterns