db-performance-watchlist
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes raw SQL commands via Prisma's
$executeRawand$queryRawmethods to monitor performance and configure database settings, such as modifyinglog_min_duration_statementinscripts/detect-slow-queries.ts. - [DATA_EXPOSURE]: The monitoring scripts access sensitive system-level views including
pg_stat_statementsandpg_stat_activity. These views contain query strings and session information that may expose sensitive application data if literals were included in database queries. - [PRIVILEGE_ESCALATION]: The skill utilizes administrative database commands (
ALTER DATABASE SET ...) which require elevated database privileges (typically superuser or database owner) to modify the database configuration at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and outputs untrusted data from the database's query history, creating a vulnerability surface where malicious queries executed against the database could influence the agent's monitoring output.
- Ingestion points: The
detectSlowQueriesfunction inscripts/detect-slow-queries.tsreads query text directly from thepg_stat_statementssystem table. - Boundary markers: None identified for the display of query text.
- Capability inventory: The skill can execute arbitrary SQL commands and generate alerts based on the ingested data.
- Sanitization: The skill applies basic truncation to query strings (
substring(0, 80)) before logging them to the console.
Audit Metadata