| name | authoring-workflows |
| description | 当你在 kimi-code 下想写 / 调试 / 启动 cc-master 的 workflow 脚本、或想套用 Claude Code dynamic-workflow API 时用——这是 kimi-code adapter 的 unsupported stub。Triggers: 在 kimi-code 里提到 Workflow 工具、`agent()` / `parallel()` / `pipeline()` / `phase()` / `workflow()`、后台 task id、workflow resume/cache/budget、"能不能把 Claude Code workflow 直接搬到 kimi"。Do NOT use when 只是做普通 kimi 任务规划、普通 Bash 脚本、或使用已验证的 kimi 后台并行能力（内置 subagent 角色 coder/explore/plan/general + Agent Swarm + 后台 Bash 任务）；本 stub 只负责阻止把 Claude Code Workflow API 当作 kimi 可用机制。启用真正 kimi 版前必须验证 deterministic multi-agent workflow execution、后台执行与完成通知、resume/cache 语义、脚本 runtime/API 形状、资源上限与 budget 报告。 |
Workflow Authoring Is Not Ported To kimi-code Yet
Do not use the Claude Code Workflow API instructions in kimi-code.
This skill is intentionally a stub in the kimi-code projection because the source skill teaches Claude Code dynamic workflows: Workflow tool, agent(), parallel(), pipeline(), phase(), log(), workflow(), background task IDs, and <task-notification> completion. kimi has no verified equivalent (no Workflow tool; the manifest has no agents field for custom roles).
Before enabling this skill for kimi-code, verify and document a kimi-native equivalent for:
- deterministic multi-agent workflow execution;
- background execution and completion notification;
- resume/cache semantics;
- supported scripting runtime and API shape;
- resource caps and budget reporting.
Until then, use built-in Task subagent roles (coder/explore/plan/general), Agent Swarm, or background Bash task dispatch per master-orchestrator-guide rather than pretending the Claude Code Workflow tool exists.
