profiling-statement-fingerprints
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill enables an agent to ingest and analyze performance data from
crdb_internal.statement_statistics, which stores SQL query texts and metadata. This data is potentially influenced by any user or application executing queries on the cluster, creating a surface for indirect prompt injection. \n - Ingestion points: SQL query templates in
SKILL.mdandreferences/sql-query-variations.mdretrieve data from thecrdb_internal.statement_statisticstable. \n - Boundary markers: The provided SQL templates and documentation do not implement boundary markers or instructions for the agent to ignore potentially malicious instructions embedded in retrieved query strings. \n
- Capability inventory: The skill allows the agent to execute SQL queries. It does not include capabilities for arbitrary command execution, file system modifications, or network exfiltration. \n
- Sanitization: There is no evidence of sanitization or filtering of the SQL query text or metadata fields before they are processed by the agent.
Audit Metadata