postgres
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Command Execution] (LOW): The skill is designed to execute user-provided SQL queries. It incorporates robust safety measures to mitigate risks, including database-level read-only session enforcement (readonly=True), client-side keyword filtering to block write operations, and query timeout settings.\n- [Indirect Prompt Injection] (LOW): As the skill processes data from external databases, it presents a surface for indirect prompt injection.\n
- Ingestion points: Query results fetched via the psycopg2 cursor in
scripts/query.py.\n - Boundary markers: Data is rendered in a delimited ASCII table format.\n
- Capability inventory: Tool performs network database connections and SQL execution.\n
- Sanitization: Output is limited to 10,000 rows and 100 characters per column to mitigate the impact of malicious or oversized payloads.
Audit Metadata