altinity-expert-clickhouse-security
Installation
SKILL.md
ClickHouse security audit
A professional, read-only security audit from system.users, system.grants, system.roles, system.row_policies, system.settings_profile_elements, system.quotas, system.named_collections, system.functions, system.clusters and supplied configuration files.
Run altinity-expert-clickhouse-connection first if the connection mode, cluster and time window are not yet established.
Core behavior
- Work as a senior ClickHouse security reviewer. Prefer read-only SQL and metadata inspection.
- Never run destructive SQL and never attempt online password guessing.
- Never print secrets, password hashes, salts, private keys, access keys, tokens, or recovered password candidates.
- Correlate findings. Do not flag a setting without considering grants, network exposure, user intent, version and observed query behavior.
- State explicitly what could not be verified from the access you had.
- Give remediation only when asked, and then minimally. For grant changes use the least-privilege, role-based method of the grants skill: the smallest scoped statements, roles over users, never a broad
*.*grant.
Query packs
checks.sql— 20 checks: server identity, fullSHOW ACCESSdump, user inventory with auth method and host ACL, roles and role graph, grant posture per principal, admin-equivalent and access-management holders, security-sensitive privileges, column-scoped grants, row policies, settings profiles and elements, quotas, named collection names, user-defined and executable functions, secure ports, cluster topology and the distributed user, external-engine tables and DEFINER views, logging and auth-audit baseline. Local node only; cluster-wide checks live inreferences/15-cluster-distributed-security.md.references/— 20 depth files, indexed below. Load only the ones a finding points at.