agentix-ceo

SUSPICIOUS: The skill is broadly consistent with an AI team orchestration service, and its data flows mostly match official Agentix endpoints, so this is not confirmed malware. However, it combines silent local credential handling, forwarding of Anthropic/GitHub secrets to the backend, unauthenticated self-hosted API access, and an explicit autopilot mode that can take ongoing actions without per-action approval, making the overall security posture medium-high risk.

No explicit malware payload is evidenced because the artifact is documentation rather than executable code. However, the described workflow presents a high security risk due to (1) plaintext persistence of sensitive Agentix credentials in a hidden home directory (`~/.agentix/credentials`), (2) delegation of third-party secrets (Anthropic API key and GitHub token) to a remote orchestration/control plane, and (3) remote playbook ingestion that governs autonomous/supervised agent behavior. Additionally, the self-hosted “no auth required” description is a major deployment red flag that could enable unauthorized local-network control if implemented similarly.