supabase

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses curl and jq to communicate with the Supabase API. These commands are constructed using parameters provided by the user or derived from database queries.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from external Supabase tables, which is a potential surface for indirect prompt injection.
  • Ingestion points: API response data from curl calls in SKILL.md.
  • Boundary markers: Data is presented within markdown table structures.
  • Capability inventory: The skill can perform network requests (GET/POST/PATCH/DELETE) and system command execution (curl, jq).
  • Sanitization: No specific filtering or escaping of database content is implemented.
  • [SAFE]: Credential management follows best practices by using environment variables (SUPABASE_URL, SUPABASE_ANON_KEY) instead of persistent storage or hardcoded values. Additionally, the skill enforces a human-in-the-loop requirement, mandating confirmation before performing delete or update operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 08:24 AM
Security Audit — agent-trust-hub — supabase