deploy-verify
Warn
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructions direct the agent to grep sensitive environment files (
.env*) for highly privileged credentials, includingSUPABASE_SERVICE_ROLE_KEYandSENTRY_AUTH_TOKEN. This exposes full secret values to the agent's execution context. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a vulnerability surface where instructions embedded in logs or web pages could influence agent behavior.
- Ingestion points: Reads database and API logs via
supabase:query_logs(Phase 2) and navigates to live web content usingplaywright-cli(Phase 4). - Boundary markers: The instructions do not define delimiters or provide 'ignore embedded instructions' warnings for the data being processed.
- Capability inventory: The skill has the capability to execute arbitrary SQL via
supabase:execute_sqland run shell commands vianpx/ Playwright. - Sanitization: No evidence of escaping or filtering logic for external content before processing.
- [EXTERNAL_DOWNLOADS]: The skill downloads and executes packages from the npm registry using
npx, specifically@playwright/cliandlangfuse-cli. These are recognized as well-known developer tools. - [COMMAND_EXECUTION]: The skill executes various shell commands to perform health checks, including CLI tool invocations and browser-based smoke tests.
Audit Metadata