The package’s launcher automatically adds first-party status hooks to user-level Claude settings, leaving a persistent configuration change that merits a warning. Inspected hook behavior reports status events, with no confirmed content exfiltration or separate attack.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
The launcher runs when Claude Code is started through Clavarium and the chat server file exists.
Impact
This creates persistent agent integration in user settings. The inspected hook commands report status events, and no concrete content theft or other attack was established.
Mechanism
The launcher invokes a setup script that writes hooks into the user-level Claude settings file; the hooks call Parley with the chat directory.
Evidence
config/launch/claude-code.shconfig/launch/claude-pulse.mjs~/.claude/settings.json