frontend-visual-qa
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from audited websites.
- Ingestion points:
scripts/visual_layout_audit.mjsandscripts/attention_inventory.mjsread rendered text and DOM structure from external URLs. - Boundary markers: Present; instructions and scripts enforce data hashing and URL redaction to prevent accidental interpretation of page content as instructions.
- Capability inventory: Browser automation and screenshot capture in
scripts/visual_layout_audit.mjs; network GET requests inscripts/verify_delivery.py. - Sanitization: Present;
redactTargetValuesDeepinscripts/visual_layout_audit.mjsfilters out sensitive information from final reports. - [COMMAND_EXECUTION]: The skill uses subprocess calls within its test suite (e.g.,
tests/test_verify_delivery.pyandtests/test_result_consumption_fixture.py) to verify script outputs. These are benign and intended only for testing CLI tool functionality. - [DYNAMIC_EXECUTION]: Browser-side scripts are executed via
page.evaluateto measure layout and typography. Additionally, the skill dynamically loads Playwright from the local project environment using predictable paths (e.g., inscripts/silent_degradation_probe.mjs). These behaviors are necessary for the skill's primary purpose and are implemented with safety in mind.
Audit Metadata