workflow-step-impl
Temporal+LangGraph の工程(step)追加/修正の実装テンプレ。Activity 冪等性、output_path/output_digest、承認待ち、監査ログを含む。
When & Why to Use This Skill
This Claude skill provides a robust implementation template for adding or modifying workflow steps within Temporal and LangGraph environments. It streamlines the development of complex distributed systems by providing skeletons for idempotent activities, output digest management (SHA-256), human-in-the-loop approval gates, and comprehensive audit logging, ensuring high reliability and traceability in agentic workflows.
Use Cases
- Scenario 1: Rapidly extending AI agent capabilities by adding new functional nodes to a LangGraph orchestration with standardized input/output handling.
- Scenario 2: Implementing resilient Temporal activities that use idempotency logic to prevent redundant processing and ensure system stability during retries.
- Scenario 3: Integrating manual approval checkpoints into automated business processes, allowing the workflow to pause and wait for external signals.
- Scenario 4: Standardizing developer workflows by using pre-defined skeletons for activity and node implementation, ensuring consistent database recording and artifact storage.
| name | workflow-step-impl |
|---|---|
| description | Temporal+LangGraph の工程(step)追加/修正の実装テンプレ。Activity 冪等性、output_path/output_digest、承認待ち、監査ログを含む。 |
使いどころ(トリガー例)
- 「工程を追加したい」「stepX を実装したい」「Activity を増やす」「LangGraph ノード追加」
手順(最小)
- 仕様書で工程ID/入出力/承認ポイントを確定(
仕様書/) - 入力の正規化と
input_digest(sha256)を定義 - Activity 実装(冪等:既存
output_pathがあれば再計算しない) - 成果物は storage に保存し、返すのは参照のみ(
output_path/output_digest/summary/metrics) - Temporal Workflow に組み込み(工程3後は signal 待機)
- DB(
step_executions/generated_files/audit_logs)の記録を追加 - レビューUI/ファイル取得APIが必要なら同時に更新
テンプレ
templates/activity_skeleton.pytemplates/step_node_skeleton.py