supabase-policy-guardrails

Warn

Audited by Gen Agent Trust Hub on Apr 23, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes Bash(supabase:*), Bash(psql:*), and Bash(npx:*) to perform database migrations, execute administrative queries, and run runtime scripts such as security audits and cost monitors.
  • [DATA_EXFILTRATION]: The cost monitoring script (scripts/supabase-cost-monitor.ts) fetches usage metrics from the Supabase Management API and transmits this data to external services via Slack webhooks.
  • [COMMAND_EXECUTION]: The security audit logic in scripts/supabase-security-audit.ts relies on a custom run_sql RPC in the database, which functions as a dynamic execution sink for SQL commands generated by the script.
  • [PROMPT_INJECTION]: The skill processes untrusted data from database schema metadata (e.g., table and column names) and local migration files, creating a surface for indirect prompt injection if those objects are maliciously named.
  • Ingestion points: supabase/migrations/*.sql files and outputs from supabase db query or psql tools.
  • Boundary markers: None present.
  • Capability inventory: Full file access (Read, Write, Edit), network operations (fetch), and subprocess execution (supabase, psql, tsx).
  • Sanitization: The provided PL/pgSQL templates use EXECUTE format(...) to escape database identifiers, which is a standard security practice for dynamic SQL construction.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 23, 2026, 08:18 PM