improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches configuration and rendering libraries from Tailwind CSS and jsDelivr (well-known services) when generating the architecture report.
- [COMMAND_EXECUTION]: Uses shell commands like
git logto identify project hotspots and platform-specific commands (open,start,xdg-open) to automatically display the generated HTML report to the user. - [DYNAMIC_EXECUTION]: Generates a self-contained HTML file in the system's temporary directory. The report configures Mermaid diagrams with
securityLevel: "loose", which enables advanced diagramming features but reduces content isolation within the browser. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted codebase files, commit history, and existing documentation to formulate architecture recommendations.
- Ingestion points: Reads git commit messages, source code files,
CONTEXT.md, and Architectural Decision Records (ADRs). - Boundary markers: None identified to distinguish codebase data from the agent's internal instructions.
- Capability inventory: Ability to write files to temporary directories, execute shell commands to open browsers, and update local project documentation via tool calls.
- Sanitization: No explicit sanitization or escaping of ingested codebase content is mentioned before it is included in the HTML report or Mermaid diagrams.
Audit Metadata