expo-smoke-test
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the project's PRODUCT.md file and runtime logs to determine test steps.\n
- Ingestion points: The skill reads PRODUCT.md, package.json, and runtime application logs (JavaScript errors, native crashes, network output) to drive automation.\n
- Boundary markers: The instructions explicitly mandate user confirmation for journeys derived from PRODUCT.md if the logic is not clear, and limit interactions with the file to read-only access. It also defines specific gates for runtime readiness and initial state.\n
- Capability inventory: The agent executes shell commands via agent-device and expo CLI tools, spawns isolated subagents, and writes documentation files to the docs/ directory.\n
- Sanitization: No specific filtering for terminal escape sequences or injection patterns in logs is described, though standard tool outputs are used and supervised by the agent coordinator.
Audit Metadata