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。
workflow-authoring【效率与协作】Author, inspect, validate, and troubleshoot Solace Agent Mesh workflow YAML/configs without starting live services.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when the task is to author, inspect, dry-validate, or troubleshoot a Solace Agent Mesh (SAM) workflow configuration that uses app_module: solace_agent_mesh.workflow.app and an app_config.workflow DAG.
Do not use this sub-skill for:
project-bootstrap.runtime-operations.sam eval suites; route to evaluation.Locate the workflow app config:
apps[] with app_module: solace_agent_mesh.workflow.app.app_config object.Check fields and examples in references/workflow-schema.md.
Check data/template behavior in references/template-resolution.md.
Run the bundled dry validator before any live run:
python sub-skills/workflow-authoring/scripts/validate_workflow_config.py path/to/workflow.yaml
Add --json for machine-readable output, --app WorkflowName to select one workflow app in a multi-app file, or --self-test to exercise the validator on an embedded tiny fixture.
If validation fails or behavior is confusing, use references/troubleshooting.md to map the symptom to a concrete config fix.
sam run, task submission, REST gateway checks, and UI verification outside this sub-skill.app_config.name is present; it becomes the workflow agent identity.app_config.namespace is present for A2A topics.workflow.description, workflow.nodes, and workflow.output_mapping are present.id and one of the supported type values: agent, workflow, switch, map, or loop.depends_on/dependencies, switch branch, map target, loop target, and exit handler points to an existing node.depends_on; otherwise branch handlers can run too early.items, withParam, or withItems.30s, 5m, 1h, 1d, or numeric seconds.input, instruction, conditions, and output_mapping match the producing node outputs.This sub-skill was distilled from SAM workflow documentation, example workflow YAML files, workflow implementation modules, workflow unit tests for Pydantic models, DAG logic, template resolution, safe conditional evaluation, and map/loop workflow target behavior. The runtime files are self-contained and do not require opening those source files.