improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill generates a local HTML architectural review and invokes platform-native commands (xdg-open on Linux, open on macOS, and start on Windows) to launch the file in a browser for user review.
- [EXTERNAL_DOWNLOADS]: The generated HTML reports include script and style references to well-known CDNs (cdn.tailwindcss.com and cdn.jsdelivr.net) to provide layout and diagramming capabilities.
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing untrusted codebase data.
- Ingestion points: Reads project source code, CONTEXT.md, and docs/adr/ files via the Explore subagent.
- Boundary markers: The instructions do not define delimiters to isolate external data from the report generation logic.
- Capability inventory: The skill can write files to the temporary directory and execute system commands to open them; the generated HTML report uses securityLevel: 'loose' for Mermaid diagrams, which allows rendering of embedded HTML tags.
- Sanitization: No specific escaping or validation is performed on the ingested code or documentation content before it is rendered in the HTML report.
Audit Metadata