improve-codebase-architecture

Warn

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The instructions in HTML-REPORT.md mandate that the Mermaid diagramming library be initialized with securityLevel: "loose". This specific configuration enables the execution of JavaScript within diagrams. Because the skill populates these diagrams with data extracted from the analyzed codebase (including file paths, module names, and commit metadata), this creates a cross-site scripting (XSS) vulnerability. If a user runs this skill on a repository containing maliciously named files or crafted commit messages, arbitrary code could be executed in the user's browser context when the report is viewed.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the codebase, forming an attack surface for indirect prompt injection.\n
  • Ingestion points: The skill reads file contents, directory structures, and git history using git log --oneline.\n
  • Boundary markers: No specific delimiters or boundary markers are instructed for use when interpolating codebase data into the analysis or the report, nor are there instructions for the agent to disregard potential commands hidden within the source code.\n
  • Capability inventory: The skill can write files to the system temporary directory, execute shell commands to open those files, and call other agent skills for domain modeling and design tasks.\n
  • Sanitization: The skill lacks instructions for sanitizing or escaping the data gathered from the codebase before it is used in report generation or processed by sub-agents.\n- [COMMAND_EXECUTION]: The skill uses system-level commands (xdg-open, open, or start) to automatically launch the generated HTML report. While this facilitates the user workflow, it involves executing shell commands on dynamically generated file paths.\n- [EXTERNAL_DOWNLOADS]: The generated report fetches external resources from well-known services, including Tailwind CSS from cdn.tailwindcss.com and the Mermaid diagramming library from cdn.jsdelivr.net. These assets are standard for providing styling and visualization capabilities.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 30, 2026, 06:39 PM