improve-codebase-architecture

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes system commands (xdg-open, open, or start) to automatically open a generated HTML report in the user's default browser. The file path is constructed using local environment variables like $TMPDIR or %TEMP%.
  • [EXTERNAL_DOWNLOADS]: The generated report includes script tags that fetch the Tailwind CSS and Mermaid.js libraries from official CDNs at cdn.tailwindcss.com and cdn.jsdelivr.net. These are well-known technology services.
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests untrusted codebase data and renders it into an HTML file viewed by the user. 1. Ingestion points: The skill reads the domain glossary, Architecture Decision Records (ADRs), and codebase structure (SKILL.md). 2. Boundary markers: None identified to prevent the agent from obeying instructions embedded in the analyzed code. 3. Capability inventory: The skill writes files to the local file system and executes shell commands. 4. Sanitization: Not explicitly mentioned; specifically, the Mermaid.js configuration in HTML-REPORT.md uses securityLevel: "loose", which enables the rendering of HTML tags within diagrams. This configuration increases the risk of Cross-Site Scripting (XSS) if the source code being analyzed contains malicious script snippets.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 07:25 AM
Security Audit — agent-trust-hub — improve-codebase-architecture