vitest-midscene-e2e
Pass
Audited by Gen Agent Trust Hub on Mar 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads a project boilerplate from the vendor's repository at
https://github.com/web-infra-dev/midscene-example.git. This is used to scaffold the testing environment and is a standard vendor-provided resource. - [COMMAND_EXECUTION]: Executes a local shell script (
scripts/clone-boilerplate.sh) which usesgitto manage the local copy of the boilerplate repository. - [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) because it processes natural language instructions to interact with third-party UIs via the
aiActAPI. - Ingestion points: User-provided task prompts and content extracted from the UI of the applications being tested (Web, Android, and iOS).
- Boundary markers: No specific delimiters or instructions are used to distinguish between the user's intended commands and content parsed from the target application's interface.
- Capability inventory: Extensive UI interaction capabilities including tapping, typing, and state assertion via Playwright, ADB, and WebDriverAgent.
- Sanitization: There is no evidence of sanitization or filtering of UI content before it is processed by the AI agent to determine the next action.
Audit Metadata