postgres-pro
Fail
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Malicious URL detection. The documentation link 'https://jeffallan.github.io/claude-skills/skills/infrastructure/postgres-pro/' provided in SKILL.md is flagged by automated scanners (URLite) as blacklisted.
- [REMOTE_CODE_EXECUTION]: Flagged external resource. The skill references an external documentation site that is potentially malicious according to automated security scans, presenting a risk of serving malicious content.
- [COMMAND_EXECUTION]: Administrative shell command instructions. The skill guides the agent to perform administrative tasks using powerful shell commands such as 'pg_basebackup', 'pg_ctl', 'systemctl', and 'pg_repack' found in references/replication.md and references/maintenance.md.
- [INDIRECT_PROMPT_INJECTION]: Vulnerability surface for indirect prompt injection.
- Ingestion points: Data enters the context via database monitoring views like 'pg_stat_statements' and 'pg_stat_activity'.
- Boundary markers: No delimiters are identified in the provided instructions to isolate untrusted data from the database.
- Capability inventory: Significant capabilities including arbitrary SQL execution and system-level shell command execution.
- Sanitization: No specific input sanitization or data validation steps are mentioned for information retrieved from the database.
Recommendations
- AI detected serious security threats
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata