walkthrough
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
agent-browsercommand-line utility to perform automated browser actions such as setting device emulation, navigating to URLs, capturing screenshots, and retrieving console errors. These commands are necessary for the skill's functional purpose of auditing live web applications and prototypes. - [EXTERNAL_DOWNLOADS]: The skill documentation recommends installing the
agent-browserpackage from the NPM registry to enable its browser automation features. This is a standard external dependency for the skill's operational requirements. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from live websites/applications and local JSON files (
jtbd.json). While it does not include explicit boundary markers or sanitization for this content, the risk is inherent to its primary purpose of analyzing external interfaces. The skill mitigates this by applying a structured cognitive walkthrough framework to the data it processes.
Audit Metadata