The agent configuration and trust changes are real, but they are activated by the user’s init command rather than an install hook. Classify this as a warning for explicit-command agent capability, not a confirmed attack.
Static reason
No blocking static signals were detected.; previous stored version diff introduced dangerous source
Trigger
A user runs the Litmus init command for an assessment.
Impact
Agent hooks may capture assessment activity without the agent’s usual trust prompt for that folder.
Mechanism
Init invokes AI tracking setup, which installs a user-scope Claude hook or falls back to project scope and sets the folder-trust flag.
Evidence
package.jsondist/commands/init.jsdist/lib/ai-tracking.js