supabase-automation

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its interaction with external data sources.
  • Ingestion points: Untrusted data enters the agent context through tools like SUPABASE_SELECT_FROM_TABLE and SUPABASE_LIST_MEMBERS_OF_AN_ORGANIZATION (found in SKILL.md).
  • Boundary markers: There are no explicit instructions for the agent to use delimiters or to ignore embedded instructions within the retrieved data.
  • Capability inventory: The skill includes the SUPABASE_BETA_RUN_SQL_QUERY tool, which allows for the execution of arbitrary PostgreSQL statements (found in SKILL.md).
  • Sanitization: No sanitization, validation, or escaping logic is defined for the content processed from external queries.
  • [DATA_EXFILTRATION]: The skill interacts with sensitive information, including service-role API keys via the SUPABASE_GET_PROJECT_API_KEYS tool. While it exposes these secrets to the agent, it correctly includes instructions to mask or truncate these values to prevent them from being leaked in the final output.
  • [COMMAND_EXECUTION]: The skill enables high-privilege database operations through the SUPABASE_BETA_RUN_SQL_QUERY tool. Although this is the primary purpose of the skill, the ability to run arbitrary SQL poses a risk if the agent's instructions are influenced by malicious data retrieved during earlier steps in a workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 07:00 AM