The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its data ingestion mechanisms.
Ingestion points: Loads and parses external project files including docs/project/tech-stack.md, prisma/schema.prisma, package.json, and .env.example. It also uses grep to search through the entire codebase.
Boundary markers: No explicit delimiters or 'ignore' instructions are used when interpolating the content of these files into the agent's context.
Capability inventory: The skill utilizes readFile, readJSON, grep, and fetch capabilities.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external files before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: Communicates with a well-known service to verify technical data.
Evidence: Fetches package metadata from the official npm registry at https://registry.npmjs.org/. This is used to confirm the existence and versions of suggested dependencies.

hallucination-detector