Caveat verdict
multi-agent-orchestrator
multi-agent-config-manager
Multi-agent orchestration framework for goal-driven research within OpenClaw; operates via local agent sessions and file-based communication without external credential exfiltration.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Findings (7)
Recursive delete from root or home — destructive command
AUDIT_v7.2.0.md · code · rm -rf ~
Dynamic import() — loads module at runtime
AUDIT_v7.2.0.md · code · import('
Uses eval() — can execute arbitrary code
lib/modelSelector.js · prose · downgraded · eval (
Uses spawn() — can execute external programs
references/protocols.md · code · spawn(
References agent configuration files
FIX_v7.2.0.md · code · AgentConfig
Uses exec() — may execute shell commands
lib/memoryPool.js · prose · downgraded · exec(
Opens WebSocket connection
CHANGES_v8.1.md · prose · downgraded · WebSocket
Permissions & capabilities
No declared permissions — minimal attack surface.
Is this flag fair?
Thanks — recorded.