improve-codebase-architecture
Warn
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (xdg-open, open, or start) to automatically open a generated HTML architectural report. While this is used for user feedback, it involves executing system commands with paths resolved from environment variables.
- [REMOTE_CODE_EXECUTION]: The architectural report template configures the Mermaid diagramming engine with "securityLevel": "loose". This setting disables security sanitization in the renderer. Since the report includes content derived from the codebase (such as file paths and module names), a malicious codebase could inject XSS payloads that execute in the user's browser context when the report is opened.
- [EXTERNAL_DOWNLOADS]: The generated reports reference resources from well-known services, including Tailwind CSS and the Mermaid diagramming library via CDN links.
- [PROMPT_INJECTION]: The skill processes untrusted input from the codebase and architectural documentation, presenting an indirect injection surface.
- Ingestion points: Codebase source files, CONTEXT.md, and documentation in docs/adr/.
- Boundary markers: No specific markers or delimiters are used to wrap codebase content to prevent it from being interpreted as instructions.
- Capability inventory: The skill can perform shell execution, file system writes, and spawn sub-agents.
- Sanitization: There is no evidence of sanitization for codebase-derived strings before they are embedded in the final HTML report.
Audit Metadata