postgresql
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external documentation fetched from third-party sources as described in references/EVIDENCE.md.
- Ingestion points: The script scripts/refresh-corpus.ts fetches content from various technology providers and saves it to the .context/postgresql-corpus directory.
- Boundary markers: The file references/EVIDENCE.md provides explicit instructions to the agent to treat fetched prose as untrusted and to never follow instructions embedded in an article, code block, or issue.
- Capability inventory: The skill instructions enable generating SQL, performing DDL changes, and suggesting operational commands across multiple reference files.
- Sanitization: The refresh script includes an extractArticleText function that removes script and style tags from fetched HTML content to provide basic structural cleaning.
- [EXTERNAL_DOWNLOADS]: The script scripts/refresh-corpus.ts fetches documentation and metadata from several well-known technology services.
- Evidence: The script targets official domains including planetscale.com, supabase.com, neon.com, databricks.com, orm.drizzle.team, and api.github.com.
- Credential Handling: The script utilizes environment variables for authenticated GitHub API requests and implements a hostname check to ensure tokens are only sent to api.github.com.
Audit Metadata