game-ui-workflow

Warn

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands using user-controlled input. In the 'Initialization' section, it calls python3 scripts/init_project.py <project-id> and in the 'Validate' section, it calls python3 scripts/validate_project.py specs/<project-id>. While the instructions advise the agent to generate IDs using only lowercase letters, numbers, and hyphens, there are no programmatic safeguards preventing the injection of shell metacharacters if the underlying scripts do not handle arguments securely.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8).
  • Ingestion points: The skill ingests untrusted data from user requests, design documents (GDD/PRD), and external reference images (mentioned in the 'Initialization' and 'Research' sections).
  • Boundary markers: The instructions implement manual 'human-in-the-loop' gates, requiring user approval for documents and assets before proceeding to subsequent stages, which acts as a partial mitigation.
  • Capability inventory: The skill possesses file-system write capabilities (creating directories, writing .md, .yaml, and .json files) and local command execution via the scripts/ directory.
  • Sanitization: The skill lacks explicit instructions for sanitizing or escaping content extracted from user requests or design documents before they are used to generate prompts for other tools or scripts.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 21, 2026, 03:10 PM
Security Audit — agent-trust-hub — game-ui-workflow