skills/impertio-studio/postgresql-claude-skill-package/postgres-impl-listen-notify/Gen Agent Trust Hub
postgres-impl-listen-notify
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely instructional and does not contain any executable scripts or hidden commands.
- [SAFE]: All external links point to official PostgreSQL documentation (postgresql.org), which is a well-known and trusted service.
- [DATA_EXPOSURE]: Code examples appropriately use environment variables for database credentials (e.g.,
process.env.DIRECT_DB_URL), following standard security practices to avoid credential exposure. - [INDIRECT_PROMPT_INJECTION]: Analyzed the risk of data poisoning from database notifications. The skill mitigates this by instructing developers to use notifications as wakeup signals only and to fetch data using parameterized queries (documented in
references/examples.md). - Ingestion points: Client-side listeners in
references/examples.mdprocessingmsg.payload. - Boundary markers: Explicit advice to treat notifications as wakeup hints and fetch row data separately.
- Capability inventory: Database read operations via standard drivers.
- Sanitization: The examples demonstrate the use of parameterized queries (
WHERE id = $1) to prevent injection from potentially malicious database payloads. - [UNVERIFIABLE_DEPENDENCIES]: References standard, well-known libraries for PostgreSQL interaction (
pgfor Node.js andpsycopgfor Python). No suspicious or unknown packages are requested.
Audit Metadata