prd-to-ux

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICAL
Full Analysis
  • [SAFE] (SAFE): No malicious patterns detected.
  • The skill uses strong instructional language ('The Iron Law', 'Not negotiable') to enforce a specific logical workflow for UX design. This is standard behavioral steering and does not target safety filters or attempt to override core agent instructions.
  • All file operations are localized. The skill reads from and writes to the current directory based on user-provided PRD files. No network activity or hardcoded credentials were identified.
  • The automated scanner finding for 'requirements.md' is disregarded as a false positive. In this context, it is a placeholder string used for documentation and file naming logic, with no associated remote download or execution pattern.\n- [Indirect Prompt Injection] (SAFE): The skill possesses a surface for indirect prompt injection, but the risk is assessed as SAFE given the limited capabilities.
  • Ingestion points: Reads from PRD files provided by the user (e.g., 'PRD.md', 'requirements.md') via the filesystem.
  • Boundary markers: Absent; the skill does not wrap input data in protective delimiters or warnings.
  • Capability inventory: Limited to writing markdown text files to the local file system. No code execution, process spawning, or network exfiltration capabilities are present.
  • Sanitization: None detected; the skill treats PRD content as raw text for analysis.
  • Assessment: Because the agent's only capability is generating a text-based specification, an embedded instruction in a PRD file cannot cause significant harm to the system or exfiltrate data.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 17, 2026, 06:45 PM