team-visual-a11y
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a structured workflow for accessibility QA including scope analysis, task dispatching, and worker monitoring. All tool permissions, such as file editing and browser interaction, are strictly aligned with its stated mission of identifying and fixing accessibility violations.
- [PROMPT_INJECTION]: An indirect prompt injection surface is present as the skill processes untrusted content from external URLs and local source files while maintaining write access to the project.
- Ingestion points: The color-auditor, typo-auditor, and focus-auditor roles navigate to external URLs and read project source files.
- Boundary markers: While the skill uses structured reporting and role-based instructions, it does not explicitly employ safety delimiters or instructions to ignore embedded commands within the processed HTML/CSS content.
- Capability inventory: The fix-implementer role has write and edit access to the filesystem (mcp__maestro__write_file, mcp__maestro__edit_file).
- Sanitization: The risk is mitigated by the auditors' use of structured data extraction (e.g., getComputedStyle) rather than processing arbitrary text prompts from the target pages.
Audit Metadata