run-flow-guide
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python utility located at
scripts/flow-guideto manage test data and generate reports. These commands are integral to the skill's state management and reporting functions. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from application UIs and guide files, creating a potential surface for indirect prompt injection. Ingestion points: The agent reads test case definitions from
.skills/<CODE>/flow-guide.jsonand application state from the browser UI. Boundary markers: The skill enforces the 'Iron Law,' requiring the agent to record quoted UI evidence and perform independent backend probes rather than relying on visual content or embedded instructions. Capability inventory: The agent is authorized to execute local CLI commands and drive a browser via Playwright or similar bridges. Sanitization: The skill relies on formatting requirements such as mandatory quoting of UI content but does not specify automated escaping or sanitization of input data.
Audit Metadata