improve-codebase-architecture
Warn
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill possesses an indirect injection surface by ingesting untrusted data from the codebase, including source files and git logs, and rendering them in an HTML report. Evidence Chain: 1. Ingestion: codebase files and git history via Explore tool. 2. Boundary markers: Absent. 3. Capability: Writing to local files and triggering browser execution. 4. Sanitization: Absent.
- [COMMAND_EXECUTION]: The skill executes platform-specific shell commands including xdg-open, open, and start to automatically open the generated report files in the user's default browser application.
- [EXTERNAL_DOWNLOADS]: The generated HTML report fetches remote assets from cdn.tailwindcss.com and cdn.jsdelivr.net to support layout and diagram rendering. These are recognized as well-known technology service providers.
- [REMOTE_CODE_EXECUTION]: The Mermaid diagramming library is initialized with a loose security level in the report scaffold. This configuration allows for the execution of active content like scripts within diagrams, posing a risk of browser-based code execution if the analyzed codebase contains malicious diagram definitions or markers.
Audit Metadata