sync-gbrain

SUSPICIOUS: the core gbrain sync and CLAUDE.md guidance update are coherent with the stated purpose, and data flow stays local/same-tooling. However, the skill bundles extra repo-admin behaviors, executes same-org but weakly pinned helper CLIs via Bun, and can auto-apply changes in spawned sessions. This looks more like a high-trust internal automation skill than outright malware, but its install/execution trust and scope are broader than ideal.