ui-template-library

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION] (LOW): Bash commands are constructed using variables derived from file data.\n
  • Evidence: references/delete-operation.md uses rm -rf and jq with $AESTHETIC_ID, which lacks sanitization against directory traversal or command injection.\n
  • Evidence: references/save-operation.md incorporates metadata-derived names into git commit messages.\n- [PROMPT_INJECTION] (LOW): The skill processes 'aesthetic prose' to influence code generation, presenting an indirect injection surface.\n
  • Ingestion points: aesthetic.md, metadata.json, and mockup HTML.\n
  • Boundary markers: None; the agent directly interprets descriptive text.\n
  • Capability inventory: Bash (rm, mkdir, jq, git), Read, and Write tools.\n
  • Sanitization: No input sanitization is performed on external prose or metadata strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:16 PM