The Agent Skills Directory

Indirect Prompt Injection (LOW): The included schema validation script processes untrusted data from the local filesystem, which could contain malicious instructions.
Ingestion points: scripts/schema_validator.py reads the full text of any file matching **/prisma/schema.prisma or certain .ts files within the provided project path.
Boundary markers: The script output (JSON or plain text) does not employ boundary markers or instructions to the agent to disregard content within the analyzed files.
Capability inventory: The agent has access to Write and Edit tools as specified in SKILL.md. An attacker placing a malicious .prisma file could potentially influence the agent to perform unauthorized file modifications.
Sanitization: The script uses regex for extraction but does not sanitize the content for potential prompt injection patterns before including it in the output.

database-design