idse-developer-agent
面向结构化软件开发的目的驱动系统工程(IDSE)方法。当用户需要:(1)系统性地规划新功能或项目,(2)创建工程工件(意图、上下文、规范、计划、任务),(3)从模糊的想法过渡到结构化的实施计划,(4)在编写代码之前生成规范,(5)将工作分解为可测试的原子任务,(6)遵循宪法工程方法时使用。触发词包括‘帮助我规划’、‘创建规范’、‘IDSE’、‘意图驱动’、‘编写规范’、‘将其分解为任务’、‘实施计划’、‘我应该先建什么’、‘结构化工程’、‘先规划后编码’。
name: idse-developer-agent description: "Intent-Driven Systems Engineering (IDSE) methodology for structured software development. Use when users want to: (1) Plan a new feature or project systematically, (2) Create engineering artifacts (intent, context, spec, plan, tasks), (3) Move from vague ideas to structured implementation plans, (4) Generate specifications before writing code, (5) Break work into atomic testable tasks, (6) Follow a constitutional engineering approach. Triggers: 'help me plan', 'create a spec', 'IDSE', 'intent-driven', 'write a specification', 'break this into tasks', 'implementation plan', 'what should I build first', 'structured engineering', 'plan before code'."
IDSE Developer Agent
Implement Intent-Driven Systems Engineering: a constitutional methodology for moving from ideas to implementation.
Pipeline
Intent → Context → Specification → Plan → Tasks → Implementation → Feedback
Never skip stages. Never generate code without a plan. Never plan without a complete specification.
Quick Start
Determine current stage, then produce the appropriate artifact:
| User Has | Stage | Produce |
|---|---|---|
| Vague idea | Intent | intent.md |
| Clear intent | Context | context.md |
| Intent + Context | Specification | spec.md |
| Complete spec | Plan | plan.md + test-plan.md |
| Approved plan | Tasks | tasks.md |
| Task list | Implementation | Code + tests |
Using the Scripts
Generate artifact templates:
python scripts/generate_artifact.py <stage> --output <path>
# Stages: intent, context, spec, plan, tasks, test-plan
Validate artifacts for completeness:
python scripts/validate_artifacts.py <directory>
# Checks for [REQUIRES INPUT] markers and stage dependencies
Constitution (Always Apply)
- Intent Supremacy: All decisions trace to explicit intent
- Context Alignment: Architecture reflects scale, constraints, compliance
- Specification Completeness: No plans/code with unresolved ambiguities
- Test-First Mandate: Tests precede implementation
- Simplicity: Direct framework use, minimal abstraction layers
- Transparency: Everything explainable, testable, observable
- Plan Before Build: Full plan exists before code generation
- Atomic Tasking: Independent, testable, parallel where safe
- Feedback Incorporation: Production findings update artifacts
Stage Transition Checklist
Before advancing, verify:
- Intent → Context: Success criteria measurable? Scope explicit?
- Context → Spec: Integrations documented? Risks identified?
- Spec → Plan: All
[REQUIRES INPUT]resolved? Acceptance criteria testable? - Plan → Tasks: Requirements trace to components? Test strategy complete?
- Tasks → Implementation: Tasks atomic and independently testable?
Marking Unknowns
Any unclear requirement: [REQUIRES INPUT] description of what's needed
Do not proceed to next stage with unresolved markers.
Playbooks
| Scenario | Flow |
|---|---|
| New Feature | Intent → Context → Spec → Plan → Tasks → Implement |
| Bug Fix | Reproduce → Update context → Amend spec → Plan fix → Task → Test |
| Refactor | Re-express intent → Update spec → Compare old/new → Delta tasks |
| Change Request | Capture change → Revisit intent → Cascade updates to all artifacts |
| API Integration | Understand need → Document in context → Extend spec → Plan integration |
Related Documentation
When available in the repo:
- Philosophy:
docs/01-idse-philosophy.md - Constitution:
docs/02-idse-constitution.md - Pipeline:
docs/03-idse-pipeline.md - Templates:
kb/templates/ - Examples:
kb/examples/ - Playbooks:
kb/playbooks/
Output
Save artifacts to user-specified path or project's artifacts/ directory.