installer
Warn
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPERSISTENCECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the
vis-networklibrary from theunpkg.comCDN and installs several MkDocs-related packages from standard registries. - [COMMAND_EXECUTION]: Setup involves executing shell commands to manage Conda environments, install packages, and deploy content to external repositories using
mkdocs gh-deploy. - [PERSISTENCE]: The skill installs global hooks into
~/.claude/settings.jsonwhich trigger monitoring scripts during every prompt submission and tool usage across all sessions. - [DATA_EXFILTRATION]: The skill sets up a persistent logging mechanism that records every user prompt and detailed execution metadata into centralized files at
~/.claude/activity-logs/. This centralized collection of prompt history creates a high-value target for data exposure. - [PRIVILEGE_ESCALATION]: Modifying global agent configuration files allows the skill to persistently monitor and influence the environment beyond the scope of a single project.
- [DYNAMIC_EXECUTION]: The visualization viewer dynamically inserts data from local JSON files into the DOM using
innerHTML, creating a potential cross-site scripting (XSS) surface.
Audit Metadata