improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
git logto identify hot spots and invokes platform-specific shell commands likexdg-open,open, orstartto automatically display the generated HTML reports in a browser.\n- [EXTERNAL_DOWNLOADS]: The generated HTML reports are configured to fetch and execute scripts from well-known Content Delivery Networks (CDNs), specificallycdn.tailwindcss.comandcdn.jsdelivr.net(for Mermaid.js).\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from the codebase and embeds it into an HTML report. Because the report uses Mermaid.js withsecurityLevel: "loose", malicious strings in the codebase could trigger cross-site scripting (XSS) when the user opens the report.\n - Ingestion points: Local codebase files,
CONTEXT.md, and architectural decision records (ADRs) within the project.\n - Boundary markers: None present; the instructions do not require the agent to sanitize or delimit codebase content before interpolation.\n
- Capability inventory: The skill can read from the file system, write to temporary directories, modify
CONTEXT.md, and execute shell commands.\n - Sanitization: Absent; the report template explicitly configures Mermaid.js to a less secure mode that permits HTML tags in nodes.
Audit Metadata