andocs
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch and install the Bun runtime using official installation scripts (
curl -fsSL https://bun.com/install | bashfor macOS and a PowerShell script for Windows). It also usesbunxto execute the latest version of theandocspackage from the official registry. - [COMMAND_EXECUTION]: Executes shell commands such as
git init,bun --version, andbunx andocs@latest serveto manage local project state and provide live previews of the documentation. It also executes Node.js build scripts (build-shared-css.mjs,build-shared-js.mjs) provided within the project repository. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process external Markdown documents, HTML prototype files, and configuration files (
prototype.json,releases.json). These files act as ingestion points for untrusted data. The skill possesses extensive capabilities including shell command execution (bunx,node) and file system access, and does not specify explicit sanitization or boundary markers when interpolating content from these files into the agent's context. - [DYNAMIC_EXECUTION]: Provides interactive
html-previewandprototypeblocks that execute HTML, CSS, and JavaScript within sandboxed iframes. It also facilitates building and running local daemons and web applications usingpnpmandnodeto support prototype editing in OpenDesign.
Audit Metadata