improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses system commands such as
xdg-open,open, orstartto automatically display the generated HTML report to the user from the system's temporary directory. - [EXTERNAL_DOWNLOADS]: The generated visual report fetches resources from well-known services, specifically
tailwindcss.comfor layout andjsdelivr.netfor the Mermaid diagramming library. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the codebase, including source code, commit history, and domain glossaries. This content is interpolated into an HTML report where the Mermaid library is initialized with
securityLevel: 'loose'. This configuration represents a potential cross-site scripting (XSS) surface if a scanned file contains a maliciously crafted Mermaid diagram. Ingestion points: Codebase files,CONTEXT.md,git log, and Architecture Decision Records (ADRs). Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded instructions within the scanned codebase files. Capability inventory: File system read access, temporary file writing, and system command execution. Sanitization: There is no evidence of sanitization or escaping of the ingested codebase content before it is rendered in the HTML report.
Audit Metadata