capture-release-evidences-cx
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes the contents of
CHANGELOG.mdto determine browser automation steps. If this file contains malicious instructions, the agent may perform unintended actions during the browser session. - Ingestion points: Workflow Step 1 reads content from
CHANGELOG.mdto map out UI elements for testing. - Boundary markers: The instructions do not specify any delimiters or warnings to prevent the agent from following instructions embedded within the changelog data.
- Capability inventory: The skill utilizes the
browser_subagenttool for browser automation and navigation, and theWritetool to create report artifacts. - Sanitization: No sanitization or validation steps are included for the content extracted from the changelog before it is used to formulate subagent tasks.
- [COMMAND_EXECUTION]: The workflow orchestrates complex browser interactions through a subagent. While mediated, these interactions involve navigating to and interacting with local network resources (
http://192.168.0.15:20128) based on dynamically interpreted instructions.
Audit Metadata