supabase-policy-guardrails
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides automated migration reviews and security audits that ingest untrusted external data in the form of SQL files, creating a surface for indirect prompt injection.
- Ingestion points: The
scripts/supabase-pre-commit.shscript reads staged migration files, and the GitHub Actions workflow inreferences/ci-cost-security.mdscans all files in thesupabase/migrations/directory. - Boundary markers: None are present; the auditing logic operates directly on the raw text of migration files without specific delimiters or instruction-isolation markers.
- Capability inventory: The skill has access to powerful tools including
Bash(supabase:*)andBash(psql:*). Furthermore, thescripts/supabase-security-audit.tsscript uses an administrativerun_sqlRPC call to execute queries on the Supabase database. - Sanitization: Sanitization is limited to simple regex and grep-based pattern matching for destructive commands (e.g.,
DROP TABLE) and does not include full SQL validation or lexical analysis of the untrusted migration content.
Audit Metadata