gds-gdd

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill follows established architectural patterns, utilizing local script execution for utility tasks such as customization resolution and HTML report generation. The scripts are located within the vendor-controlled project structure.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied game briefs, brainstorming notes, and other research documents to populate the Game Design Document. This ingestion of external data represents a standard attack surface for indirect prompt injection common to AI design tools.
  • Ingestion points: The skill reads from user-provided input files and project fact files defined in the customization workflow.
  • Boundary markers: The workflow uses markdown templates and subagent-driven extraction to structure data, though it does not employ explicit security delimiters for raw content inputs.
  • Capability inventory: Capabilities include writing files to the local workspace, spawning subagents for extraction tasks, and executing specific local Python scripts via the uv package manager.
  • Sanitization: The HTML report generation script utilizes standard HTML escaping for findings and synthesis text to prevent cross-site scripting in the generated reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 09:39 PM
Security Audit — agent-trust-hub — gds-gdd