开源项目 Harness 工作流 YAML 生成器
为 AI 编码项目生成可复用的 YAML 工作流定义,让 AI 编码过程可重复、可审计
You are an expert in AI-assisted software development workflows. Generate a complete YAML workflow definition for the following development task. Task: [describe the feature/fix/refactor] Repo: [language, framework, test setup] Agent: [Claude Code / Codex / Goose / other] Generate a workflow YAML with these phases: ```yaml name: <descriptive-name> version: 1 phases: - id: explore type: ai prompt: | Explore the codebase structure. Identify: - Entry points relevant to the task - Existing patterns and conventions - Test infrastructure and coverage - Dependencies that may be affected artifacts: [exploration-report.md] - id: plan type: ai depends_on: [explore] prompt: | Based on the exploration report, create a detailed implementation plan. - Break into atomic, testable steps - Identify risks and edge cases - Estimate complexity per step artifacts: [implementation-plan.md] gate: human-approval - id: implement type: ai-loop depends_on: [plan] max_iterations: 10 prompt: | Implement the next incomplete task from the plan. After each change, run the relevant tests. exit_condition: all-tasks-complete fresh_context: true - id: validate type: deterministic depends_on: [implement] commands: - npm test - npm run lint - npm run typecheck - id: review type: ai depends_on: [validate] prompt: | Review all changes made. Check for: - Code quality and consistency - Missing tests - Security concerns - Performance implications artifacts: [review-report.md] - id: pr type: deterministic depends_on: [review] commands: - git add -A - gh pr create ``` Customize every prompt and command for my specific task. Add error handling and rollback steps where appropriate.
如何使用这条提示词
- 1复制上方完整提示词。
- 2在对应模型中替换主题、人物或风格变量。
- 3生成后记录有效调整,形成自己的版本。



