dbdesign
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-provided 'requirements, domain description, or interview notes' (SKILL.md). This data is then used to generate database designs and DDL.\n
- Ingestion points: Requirements dossier and interview notes provided by the user (SKILL.md).\n
- Boundary markers: The skill contains explicit instructions to treat user-provided text as evidence, not instructions, and to move any embedded commands to an 'open questions' log rather than executing them (SKILL.md).\n
- Capability inventory: The agent is instructed to execute generated DDL against an in-memory SQLite database (SKILL.md).\n
- Sanitization: The skill relies on the LLM's adherence to the 'evidence vs. instructions' rule and lacks automated sanitization or filtering of external content.\n- [DYNAMIC_EXECUTION]: The skill generates SQL DDL scripts based on user input and is instructed to validate them by executing the code against a local, in-memory SQLite database (
sqlite3 :memory:) as described in SKILL.md. This constitutes dynamic script generation and execution.\n- [COMMAND_EXECUTION]: The skill instructions direct the agent to invoke thesqlite3command-line tool to verify generated scripts (SKILL.md). While limited to an in-memory database, it involves direct execution of shell commands.
Audit Metadata