search-pg-hackers
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches mailing list threads and patch attachments from the official PostgreSQL archive at https://www.postgresql.org. This is a well-known and trusted source for PostgreSQL community data.
- [COMMAND_EXECUTION]: Uses a local Python script (scripts/postgresql_archive.py) to manage data retrieval and storage. The script utilizes the Python standard library for all operations and performs filename sanitization using Path(name).name to prevent directory traversal attacks when saving attachments.
- [PROMPT_INJECTION]: The skill handles untrusted data from the mailing list (Category 8), which is a necessary part of its function. 1. Ingestion points: Thread HTML and attachments fetched from postgresql.org. 2. Boundary markers: SKILL.md provides explicit warnings to treat all external data as untrusted and never execute attachments. 3. Capability inventory: The script can write files to a user-specified directory for the patch store. 4. Sanitization: The script uses safe_filename logic to clean all downloaded filenames.
- [SAFE]: No malicious patterns such as obfuscation, exfiltration to unauthorized domains, or privilege escalation were detected.
Audit Metadata