book-installer

Warn

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill installs and executes several shell scripts to manage infrastructure and monitoring. This includes the 'Skill Tracker' which installs persistent hooks in ~/.claude/settings.json (UserPromptSubmit, PreToolUse, PostToolUse) to monitor all agent interactions. It also includes a cover generation script that uses browser automation on macOS requiring Accessibility permissions to control the user's browser.
  • [DATA_EXFILTRATION]: The feedback.js component includes functionality to transmit telemetry, including page URLs, timestamps, and user agent strings, to an external endpoint that can be defined by the user in the mkdocs.yml configuration.
  • [EXTERNAL_DOWNLOADS]: The project templates and snippets configure the resulting textbook to load multiple third-party JavaScript libraries from public CDNs. These include vis-network (from unpkg.com), KaTeX and MathJax (from jsdelivr.net or unpkg.com), and Giscus (from giscus.app). While these are common libraries, they represent a remote script execution surface in the generated site.
  • [DATA_EXPOSURE]: The 'Skill Tracker' functionality creates a local log of every prompt the user sends to the AI agent, along with session IDs and execution metadata, in the ~/.claude/activity-logs/ directory. This results in a permanent record of all user interactions with the AI across all projects.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 27, 2026, 09:23 PM