pm-image2proto

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches the Mermaid.js library from a well-known CDN (cdnjs.cloudflare.com) to render process flows. This is a trusted service and the download is used for the skill's primary purpose.
  • [COMMAND_EXECUTION]: The skill allows for the creation and modification of HTML files in a user-defined directory (output_root). It uses file system tools to persist these prototypes based on user design requests.
  • [PROMPT_INJECTION]: The skill processes user-provided images and interaction history to generate code, creating an indirect prompt injection surface.
  • Ingestion points: UI screenshots, natural language user feedback, and the persistent interaction log file (references/learning_log.jsonl).
  • Boundary markers: None present; the skill lacks delimiters or instructions to distinguish between benign design intent and potential malicious instructions embedded within the images or logs.
  • Capability inventory: The skill has the capability to write and modify files (HTML prototypes and configuration files) on the local file system.
  • Sanitization: No validation, escaping, or sanitization is performed on the attributes extracted from screenshots or the feedback stored in the learning log before they are incorporated into code generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 02:39 AM
Security Audit — agent-trust-hub — pm-image2proto