improve-codebase-architecture

Pass

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill generates an HTML report that fetches styling and diagramming libraries (Tailwind CSS and Mermaid.js) from well-known services (cdn.tailwindcss.com and jsdelivr.net).
  • [COMMAND_EXECUTION]: To present findings to the user, the skill automatically executes operating system-specific commands (xdg-open, open, or start) to open a generated file in the temporary directory.
  • [PROMPT_INJECTION]: The skill processes untrusted repository data, creating a surface for indirect prompt injection.
  • Ingestion points: Source code files, architectural decision records (ADRs), and domain glossary files read from the local repository (SKILL.md, Step 1).
  • Boundary markers: Absent; no specific delimiters or instructions are provided to the agent to separate untrusted codebase data from its own instructions.
  • Capability inventory: The skill has the ability to read and write files and execute shell commands to open the report.
  • Sanitization: Absent; the agent is not instructed to sanitize or escape data before embedding it in the HTML report. The report configuration specifically uses securityLevel: "loose" for Mermaid diagrams, which increases the risk that malicious strings in the source code could execute script content in the user's browser.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 10, 2026, 11:49 AM
Security Audit — agent-trust-hub — improve-codebase-architecture