The Agent Skills Directory

[COMMAND_EXECUTION]: The tool SUPABASE_BETA_RUN_SQL_QUERY enables the execution of arbitrary PostgreSQL statements. This high-privilege capability is necessary for database administration but presents a risk of unauthorized data modification or destructive operations if the agent is misled or if the SQL input is not strictly validated.
[CREDENTIALS_UNSAFE]: The skill includes instructions for using SUPABASE_GET_PROJECT_API_KEYS, which returns live service-role secrets. The author provides explicit warnings to never log full values and to mask keys in output, which demonstrates security awareness but acknowledges the presence of sensitive data handling.
[EXTERNAL_DOWNLOADS]: The setup instructions direct the user to add an external MCP server at https://rube.app/mcp. This creates a dependency on third-party infrastructure that is not included in the standard list of trusted vendors.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources and possesses significant administrative capabilities.
Ingestion points: Data is retrieved via SUPABASE_SELECT_FROM_TABLE and SUPABASE_LIST_TABLES (SKILL.md).
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when the agent processes retrieved database content.
Capability inventory: The skill can execute arbitrary SQL via SUPABASE_BETA_RUN_SQL_QUERY and access sensitive credentials via SUPABASE_GET_PROJECT_API_KEYS (SKILL.md).
Sanitization: Instructions recommend masking API keys, but there are no defined validation or escaping mechanisms for general data retrieved from the database.

supabase-automation