The Agent Skills Directory

Prompt Injection (SAFE): The skill contains standard instructional content for Supabase administration without any attempts to bypass safety filters or override agent behavior.
Data Exposure & Exfiltration (SAFE): No hardcoded credentials, sensitive file paths, or suspicious network operations were identified. The use of Supabase-specific functions like auth.uid() and auth.jwt() is appropriate for the stated purpose.
Unverifiable Dependencies (SAFE): No external package installations or remote script executions are present. The skill relies on standard tools and the mcp__supabase__* protocol for its operations.
Indirect Prompt Injection (LOW): The skill is designed to process and generate SQL schemas and migrations, which are untrusted data entry points.
Ingestion points: Reads and modifies migration files in supabase/migrations/ and processes schema definitions.
Boundary markers: None explicitly defined in the prompt instructions to isolate external SQL content.
Capability inventory: Uses Bash, Write, and Edit for file system operations, and mcp__supabase__* for direct database interaction.
Sanitization: The skill provides best-practice templates (e.g., SECURITY DEFINER use, index optimization) which encourage safe output generation, though it does not explicitly sanitize input SQL.
Dynamic Execution (SAFE): There is no evidence of runtime code generation, unsafe deserialization, or library injection. SQL execution is performant and uses standard PostgreSQL patterns.

supabase-admin