design-code
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple local scripts and binaries at runtime to verify output quality.
- Evidence:
python .claude/skills/_ux-ui-shared/scripts/lint_hardcodes.py,python .claude/skills/_ux-ui-shared/scripts/validate_theme_refs.py,python .claude/skills/_ux-ui-shared/scripts/contrast.py,node .claude/skills/_ux-ui-shared/scripts/taste_audit.mjs <file>,node .claude/skills/_ux-ui-shared/scripts/verify_states.mjs <file>, andnode .claude/skills/_ux-ui-shared/scripts/accuracy_report.mjs. - Context: These tools are used for linting, accessibility checks, and design consistency within the
.claude/skills/_ux-ui-shared/directory structure. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user requests to generate code and then passes that generated code as an argument to local scripts (
<file>), creating a potential injection vector if the scripts do not properly sanitize input. - Ingestion points: User requests for UI components or screens (SKILL.md).
- Boundary markers: None specified for the script execution phase.
- Capability inventory: Python and Node.js execution via subprocess calls for linting and auditing.
- Sanitization: Not explicitly mentioned for the script arguments.
Audit Metadata