stop

SUSPICIOUS: the overall workflow mostly matches a session-stop skill, but its footprint is high risk because it relies on unverifiable atlas-session MCP tooling and can perform real-world actions like pushing code, creating PRs, and settling token bounties. The main concern is trust and data flow opacity around the atlas-session/contract backend rather than clear malicious intent.