review-pg-patch

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external data from the PostgreSQL hackers mailing list, creating an indirect prompt injection surface. Data is ingested from JSON files via scripts/review_memory.py ingest and stored in a local SQLite database. The skill instructions (references/memory-workflow.md) serve as a boundary by requiring verification of source URLs and comparison with current code. The skill's capabilities include writing local files and maintaining an SQLite store, though no automated sanitization is applied to ingested email bodies.
  • [COMMAND_EXECUTION]: The skill generates benchmark and test scaffolds, including a run.sh file with executable permissions. The generated file uses a static template intended for user completion, representing a standard development utility.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 10:38 AM
Security Audit — agent-trust-hub — review-pg-patch