builder-smoke-test
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple internal shell scripts (scaffold.sh, preflight.sh, seed-multi-user.sh, wait-for-server.sh) to prepare the hermetic testing environment and manage the local development server.
- [DYNAMIC_EXECUTION]: The skill uses templates to generate functional Mastra application code, which is then dynamically executed using the mastra dev command to facilitate integration testing.
- [CREDENTIALS_UNSAFE]: The skill handles sensitive API keys (OpenAI, WorkOS) and session cookies. It includes mechanisms for retrieving credentials from user-approved local configuration files and a specific debug route to access httpOnly session cookies for automated browser testing, all managed through documented user-consented workflows.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from API responses and browser-based UI interactions, presenting an attack surface for instructions embedded in external data. Ingestion points include API responses and UI content processed by the agent. Capability inventory includes shell command execution, project scaffolding, and browser automation. The risk is minimized by the skill's operation within a hermetic test environment.
Audit Metadata