计划执行与检查节点
executing-plans
按既定实施计划完成工作,保持过程检查和反馈。
task-yaml【效率与协作】Author, validate, troubleshoot, and adapt SkyPilot task and service YAMLs safely.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when the user needs to write, review, convert, or debug a SkyPilot task YAML, a SkyServe YAML's task portion, or Python code that constructs sky.Task and sky.Resources from YAML-compatible config.
references/yaml-reference.md for accepted top-level fields, resource forms, service YAML fields, Python construction APIs, and safe validation flow.scripts/validate_task_or_service_yaml.py for parser-only validation when SkyPilot is installed; it does not launch clusters, contact clouds, or write resources.references/troubleshooting.md when validation fails or when YAML behavior is ambiguous.name, workdir, num_nodes, resources, envs, secrets, api_server_access, volumes, file_mounts, setup, run, and config.infra, accelerators, cpus, memory, instance_type, use_spot, disk_size, ephemeral_storage, disk_tier, network_tier, ports, labels, any_of, ordered, autostop, job_recovery, priority, and priority_class.MOUNT, COPY, MOUNT_CACHED, mount-cached presets, and Kubernetes volume references.sky.Task.from_yaml(), sky.Task.from_yaml_config(), sky.Task.from_yaml_str(), and sky.Resources(...).service: section and service-spec fragments, then routing runtime SkyServe operations elsewhere.../cluster-operations/SKILL.md.../managed-jobs/SKILL.md.../serving/SKILL.md.../infrastructure-storage/SKILL.md.../sdk-api-server/SKILL.md.../repo-development/SKILL.md.Run the bundled parser helper from this sub-skill directory or with an explicit path:
python scripts/validate_task_or_service_yaml.py --kind auto path/to/task.yaml
Use --kind task for normal task YAMLs and --kind service for SkyServe YAMLs or service-spec fragments. The helper imports SkyPilot parser APIs only; it is not a dry run, optimizer call, or credential check.