using-design-skills
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to inspect the local environment and discover available tools. It uses 'jq -r ".skills|keys[]" ~/.agents/.skill-lock.json' to retrieve the list of installed skills and 'head -8' to check metadata headers within those skill files.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an orchestration workflow that processes user tasks and project-specific state, creating a surface where malicious instructions could influence the execution of sub-agents.
- Ingestion points: Ingests user-provided task descriptions and reads design tokens/constraints from 'DESIGN.md' located in the repository.
- Boundary markers: Attempts to separate instructions using specific headers (e.g., 'DESIGN CONTEXT', 'SKILL'), which provides some structure but can be bypassed by adversarial content within the processed files.
- Capability inventory: Manages persistent workers via 'agent-tmux', writes data to 'DESIGN.md', and generates complex HTML/SVG artifacts.
- Sanitization: Content from 'DESIGN.md' is interpolated into worker prompts without explicit sanitization or escaping mechanisms documented in the skill.
- [EXTERNAL_DOWNLOADS]: The skill references 'https://namethatui.com/' as an advisory vocabulary dictionary for UI components. This is a well-known service for design terminology.
Audit Metadata