web-form-fill

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a robust human-in-the-loop workflow. It requires the agent to collect fields and prepare content in the workspace for review before attempting to fill forms, ensuring user oversight of the data being processed.- [INDIRECT_PROMPT_INJECTION]: While the skill ingests external data from web pages and workspace files, it mitigates potential injection risks by instructing the agent to verify rendered values against placeholders and strictly requiring a final 'human review' of all screenshots and field lists before submission.- [COMMAND_EXECUTION]: Form interactions are performed using specialized browser tools (xiaobei_browser, act). The instructions explicitly forbid the use of evaluate to directly assign values to elements to bypass framework validation, ensuring the automation respects the site's security and state logic.- [DATA_EXFILTRATION]: The skill handles file uploads and workspace writing within the bounds of standard project requirements and media policies. It uses a session management and lease system that appears to be a platform-standard concurrency and security mechanism.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 11:03 AM