smoke
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill employs dynamic module loading in
runner/cli.mjsto execute scenario files defined during the test setup. Additionally, it provides anevalcapability within the browser handle indevices/chromium.mjsto execute JavaScript code in the context of the web page being tested, which is standard for web automation. - [COMMAND_EXECUTION]: The adapter in
devices/chromium.mjsperforms process management using system commands, specifically callingpsto verify process identity andprocess.killto terminate browser instances during cleanup. This is used to prevent resource leakage on the host machine. - [EXTERNAL_DOWNLOADS]: Fetches and installs the
playwright-corepackage and Chromium browser binaries from official registries to establish the automation environment. - [INDIRECT_PROMPT_INJECTION]: The skill's architecture creates an attack surface where malicious instructions could be embedded in the web pages being automated. These instructions could be ingested when the agent reviews test outputs.
- Ingestion points: Browser console logs, page error messages, and return values from script evaluations collected by the device adapter in
devices/chromium.mjs. - Boundary markers: None; there are no specific instructions or delimiters used to isolate test data from the agent's command logic.
- Capability inventory: File system writes for artifacts, dynamic script execution, and process lifecycle management.
- Sanitization: No filtering or sanitization is applied to the data retrieved from the browser before it is included in reports and processed by the agent.
Audit Metadata