supabase
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes a database trigger (
handle_new_user) that extracts data fromraw_user_meta_dataprovided during authentication. This creates a surface where external data enters the system context. - Ingestion points: The
public.handle_new_userfunction inSKILL.mdprocesses metadata fromauth.users. - Boundary markers: None are present in the provided SQL examples to delimit or validate this metadata.
- Capability inventory: The skill utilizes database write operations and network requests within Edge Functions.
- Sanitization: No explicit validation or sanitization of the metadata fields is demonstrated in the SQL trigger.
- [COMMAND_EXECUTION]: The skill guides the agent to use the
supabaseCLI for project initialization, local development, and deployment. These commands are standard for the platform's development workflow. - [EXTERNAL_DOWNLOADS]: The skill references downloads for the official Supabase CLI through NPM and fetches dependencies for Edge Functions from well-known services including
deno.landandesm.sh. It also demonstrates integration with the Resend API for email delivery, which is an established service.
Audit Metadata