skills/laststance/skills/visual-lint/Gen Agent Trust Hub

visual-lint

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the 'Bash' tool to manage the target application's lifecycle, executing commands such as 'kill-port 9222', 'kill-all', and 'pnpm dev' to ensure a clean state before capturing screenshots (SKILL.md).
  • [DYNAMIC_EXECUTION]: The skill performs runtime execution of JavaScript within the target application's context using the 'playwright-cli --s=default eval' command to gather environmental data like 'window.devicePixelRatio' (SKILL.md).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external visual data through a Vision-Language Model, creating a surface where UI content could attempt to influence the agent.
  • Ingestion points: Full-window and element-scoped screenshots captured via 'playwright-cli' (SKILL.md).
  • Boundary markers: Relies on specific rubric references (references/defect-rubric.md) but lacks explicit character-level delimiters for the visual input.
  • Capability inventory: Access to 'Bash' (command execution), 'Write' (report generation), and 'Agent' (sub-agent spawning) tools (SKILL.md).
  • Sanitization: No automated sanitization or filtering of screenshot content is performed; relies on process rules in references/anti-false-positive.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 09:21 AM
Security Audit — agent-trust-hub — visual-lint