The explicit, disclosed litmus init flow installs Litmus AI-capture hooks and changes coding-agent trust settings, including skipping Codex's hook review and folder-trust prompts. This is a real agent-control risk, but the source does not establish a concealed attack or credential theft.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
Explicit `litmus init` setup for an assessment.
Impact
AI activity can be captured across supported tools, and the assessment folder's trust prompt is bypassed. The setup is disclosed, but changing agent trust settings creates a broader agent-control risk.
Mechanism
The CLI writes hooks into agent configuration and pins hook trust; those hooks capture prompts and responses for the assessment and upload events to its configured backend.
Evidence
dist/commands/init.jsdist/lib/ai-tracking.jsdist/lib/hook-logger.cjs~/.codex/hooks.json~/.codex/config.toml~/.claude/settings.json~/.claude.json