working-with-mbx

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download and installation of the mr-boxington binary and the mise tool manager from GitHub repositories (github.com/jdx). These are established tools in the developer ecosystem.
  • [EXTERNAL_DOWNLOADS]: The update-docs.sh script uses npx to execute a crawler (@mdream/crawl) to fetch documentation content from https://mr-boxington.jdx.dev.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes a documentation update mechanism that creates a surface for indirect prompt injection.
  • Ingestion points: The update-docs.sh script fetches external markdown content from https://mr-boxington.jdx.dev/** and saves it to the references/ directory.
  • Boundary markers: Not present; the update script performs a direct replacement of existing reference files with newly crawled content.
  • Capability inventory: The skill is primarily informational but supports shell command execution for tool management (e.g., mbx build, mbx gc) and file system operations within the workspace target directories.
  • Sanitization: No sanitization is performed on the fetched external content before it is stored for use as agent context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 07:17 AM
Security Audit — agent-trust-hub — working-with-mbx