hud
Fail
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill establishes persistence by instructing the agent to modify
~/.claude/settings.jsonto register astatusLinecommand. This ensures the custom scriptomc-hud.mjsis executed automatically on startup and periodically during use. - [REMOTE_CODE_EXECUTION]: The generated
omc-hud.mjsscript performs dynamic code loading using asynchronousimport(). It attempts to load and execute code from paths computed at runtime, including development directories and theoh-my-claudecodenpm package. - [COMMAND_EXECUTION]: The setup procedure employs several
node -eshell commands to automate directory creation, file system verification, and the modification of file permissions to make scripts executable. - [EXTERNAL_DOWNLOADS]: The skill references and depends on the
oh-my-claudecodenpm package. The configuration script is designed to dynamically fetch and execute this package if found in the environment or registry. - [DATA_EXFILTRATION]: The instructions require the agent to read the full contents of
~/.claude/settings.json. This access exposes potentially sensitive application configuration and session data to the agent's context.
Recommendations
- AI detected serious security threats
Audit Metadata