The package exposes a real dangerous remote agent-control capability with permissive defaults, so it warrants a warning. Static source inspection does not establish malicious install behavior, stealth persistence, or exfiltration.
Static reason
No blocking static signals were detected.
Trigger
A user pairs and starts the manager; the configured AWB server then sends agent-management events.
Impact
A compromised or overly trusted configured server could cause managed agents to operate in configured working directories with bypassed Codex safeguards and seeded Claude trust.
Mechanism
Remote management of AI CLI sessions with permissive execution settings.
Evidence
dist/lib/cli-adapters/codex.jsdist/lib/cli-adapters/claude.jsdist/lib/event-stream.jsdist/lib/agent-manager-commands.js$AWB_AGENT_MANAGER_HOME/config.json$AWB_AGENT_MANAGER_HOME/agents/<agent_id>/config.json$AWB_AGENT_MANAGER_HOME/agents/<agent_id>/credential.json$AWB_AGENT_MANAGER_HOME/agents/<agent_id>/mcp-config.json