lightfast-db

Pass

Audited by Gen Agent Trust Hub on Apr 25, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill's functionality is scoped to read-only database interactions, with explicit notes preventing data modification via SQL or pnpm migration tools.
  • [PROMPT_INJECTION]: The skill processes user-supplied data, presenting an indirect prompt injection surface.
  • Ingestion points: User input is ingested through the $ARGUMENTS variable in command/lightfast-db.md.
  • Boundary markers: User input is delimited by <user-request> tags to isolate it from skill instructions.
  • Capability inventory: The skill can execute SQL queries and inspect database metadata via the postgres MCP server.
  • Sanitization: There is no automated sanitization, although the agent is instructed to perform case conversion on table names.
  • [COMMAND_EXECUTION]: The skill dynamically generates SQL queries from natural language input. This is the intended behavior for the skill, and the risk of abuse is mitigated by the read-only transaction constraint and the specific scope of the postgres MCP query tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 25, 2026, 10:50 AM
Security Audit — agent-trust-hub — lightfast-db