The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted data from local files like schema.prisma, drizzle.config.ts, and package.json to inform its logic. If these files are maliciously crafted, they could influence the agent's subsequent Write or Bash actions.
Ingestion points: reads configuration and schema files from the local filesystem (Phase 1: Entity Analysis).
Boundary markers: No specific delimiters or instructions are provided to the agent to disregard natural language instructions found within the data files.
Capability inventory: Access to Write, Edit, and Bash tools allows for significant system interaction based on processed input.
Sanitization: The skill lacks explicit sanitization or validation of the content read from files before using it in code generation prompts.
[Command Execution] (SAFE): The skill uses the Bash tool to run type checks and tests (Phase 4: Verification). This is appropriate for the stated purpose of verifying generated code and does not involve suspicious or obfuscated commands.

crud