altinity-expert-clickhouse-grants
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes diagnostic SQL queries against a ClickHouse database using a
checks.sqlfile. These queries read fromsystem.errors,system.query_log,system.grants, andsystem.role_grantsto identify recent access denied errors and current user permissions. This is standard diagnostic behavior for a database administration skill. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from database logs (exception messages and user names) to suggest SQL
GRANTstatements. - Ingestion points: Reads from
system.errorsandsystem.query_log(checks.sql). - Boundary markers: The skill uses clear section headers and code blocks to separate instructions from data patterns.
- Capability inventory: The skill generates SQL code for the user to review and execute; it does not possess autonomous write or network capabilities itself.
- Sanitization: The skill advises the agent to propose the 'smallest set' and 'narrowest form' of grants, effectively serving as a logical filter against over-privileged output derived from log data.
Audit Metadata