octo-skills

Conversational cold-start advisor for Multica squads. Use this whenever the user wants to set up, configure, bootstrap, or onboard a Multica squad / team of agents — e.g. "帮我配置一个 multica 小队", "组建一个 multica 团队", "冷启动一个 squad", "建一队 agent", "set up a multica squad", "configure my squad", "onboard a team on multica", or just describes a team goal + rough roles and wants it stood up. The user only needs to describe (1) the squad's overall goal and (2) a rough team composition / division of labor; this skill clarifies the gaps through multi-turn conversation, then designs and (after approval) provisions the complete squad via the `multica` CLI: agents (with professionally authored instructions), each agent's skills, the squad with a leader, members with semantic roles, and squad-level instructions. Also triggers for incrementally updating / adding members to an existing squad. Trigger even if the user does not say the word "skill" — if they talk about standing up a Multica team, this is the entry point.

An OpenClaw loop-engineering standard built on octoim + octospec + octo-multica-cloud. A three-ring nested contract (orchestration / execution / verification) with a loop-uncheatable acceptance gate. Use when receiving a code-implementation / bug-fix / feature request, designing or reviewing cross-agent dispatch, or auditing whether a pipeline conforms to the loop standard. Triggers on "loop process", "engineering standard", "dispatch standard", "acceptance gate", "stop condition", "fake acceptance", "loop engineering". NOT for trivial one-line fixes / typos / pure config (use the trivial escape hatch and dispatch the implementer directly).

Call cloud multica (multica.imocto.cn) from octo via user-configured octo-multica-key secret. Generic/portable; onboarding guide + runtime discovery.