ux-flows
Warn
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PRIVILEGE_ESCALATION]: The documentation in
references/browser-engine.mdinstructs the user or agent to executesudo safaridriver --enable. This command requires administrative privileges to enable remote automation for the Safari browser, which could be exploited if run without oversight. - [REMOTE_CODE_EXECUTION]: The skill frequently references running code via
npx(e.g.,npx sheleg-design-skill,npx skills add ssheleg/sheleg-design-skill). While these tools are authored by the vendorssheleg, they involve the dynamic execution of code downloaded from the npm registry at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill defines a funnel research workflow (
references/funnel-research.md) that uses browser automation (e.g., Playwright) to visit and capture data from competitor websites. - Ingestion points: Competitor websites via Playwright/WebDriver, Refero and Mobbin MCP tool outputs, and Figma design files.
- Boundary markers: No specific delimiters or instructions to ignore embedded prompts are mentioned for untrusted content.
- Capability inventory: The skill can execute Python scripts (
lint.py), run shell commands vianpx, and perform file system writes (creating UX plans indocs/ux/plans/). - Sanitization: No explicit sanitization or filtering logic is detailed for the ingested web content.
- [COMMAND_EXECUTION]: The skill requires the execution of a local Python script (
python3 docs/ux/lint.py) for project linting and various shell commands for Figma integration and project maintenance. - [EXTERNAL_DOWNLOADS]: The skill recommends installing external plugins and dependencies from npm and marketplace registries (e.g.,
task-pipeline,sheleg-design-skill) to extend its functionality.
Audit Metadata