inspect-ui
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill utilizes
npx playwright screenshotto obtain visual evidence of UI changes. This invokes the Playwright browser automation tool, which is fetched and executed from a well-known public registry at runtime.\n- [COMMAND_EXECUTION]: The agent is instructed to discover and run development commands defined within the local repository to start the application. This involves the execution of scripts found in the untrusted source code being reviewed.\n- [INDIRECT_PROMPT_INJECTION]: The skill renders untrusted frontend code (HTML, CSS, and component files) in a browser, creating a potential vector for instructions embedded in the code to affect the agent's behavior.\n - Ingestion points: Browser-rendered surfaces such as HTML, CSS, JSX, and Vue files identified in the git diff are processed.\n
- Boundary markers: The instructions do not define specific delimiters or security warnings to separate the processed code content from the agent's internal reasoning context.\n
- Capability inventory: The skill has access to shell commands (git, grep, npx), local file creation for screenshots, and full browser control.\n
- Sanitization: No validation or sanitization is performed on the rendered code or the generated visual data prior to analysis.
Audit Metadata