design-system

Pass

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a 'Reference Input' phase (Phase 0) that is vulnerable to indirect prompt injection.\n
  • Ingestion points: Untrusted data enters the agent context via user-provided folder paths, individual files, and external website URLs processed through the WebFetch tool as described in SKILL.md.\n
  • Boundary markers: The skill does not define explicit delimiters or instructions to ignore potential commands embedded in the fetched content or image metadata during its analysis protocol.\n
  • Capability inventory: The skill possesses file-write capabilities to both local and user-level directories (~/.claude/.context/design-systems/) and can trigger external image generation using the creator-stack:nanobanana skill.\n
  • Sanitization: There is no documentation of sanitization or validation of the text or visual properties extracted from external references.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes a WebFetch mechanism to retrieve content from arbitrary URLs provided by the user for the purpose of visual design analysis. This is a core feature but represents a point of interaction with external, untrusted infrastructure.\n- [COMMAND_EXECUTION]: While the skill manages complex file system organizations and interacts with other agent skills, it does not invoke arbitrary system shell commands or execute unverified local scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 7, 2026, 03:28 AM