agentsclimarketplace

Nodeflow architecture

Skill CAI5946/NodeFlow/skills/nodeflow-architecture

Use after docs/PRD.md exists to decide architecture and create ARCH, CONTEXT, and DECISIONS docs before task breakdown.From its SKILL.md

Install
npx -y skills add CAI5946/NodeFlow --skill nodeflow-architecture

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.0 KB, as published. Nobody here has run it

Nodeflow Architecture

Node Metadata

  • Node ID: nodeflow-architecture
  • 作用: 生成架构、领域词典、决策记录,并触发项目级 agent 规则
  • 输入: docs/PRD.md、可选 docs/BRIEF.md, docs/NODEFLOW.md, docs/ASSUMPTIONS.md
  • 输出: docs/ARCH.md, docs/CONTEXT.md, docs/DECISIONS.md, AGENTS.md, CLAUDE.md
  • 前置条件: docs/PRD.md 已存在且核心需求明确
  • Gate: 技术栈、模块边界、共享接口或数据模型不清时停止
  • 默认 checkpoint: Y
  • 可跳过条件: 已有已确认 docs/ARCH.md 和 AGENTS.md
  • 下游节点: nodeflow-task-breakdown
  • 适用风险: High

Node Status

  • Status: core
  • Role: architecture node
  • Replacement: none

Core Rules

  • Reply in Chinese.
  • Stay in architecture and project-rule preparation.
  • Do not implement code or generate TASKS.md.
  • Read docs/PRD.md before deciding architecture.
  • Generate AGENTS.md only after architecture is defined enough to name tech stack, structure, commands, and validation expectations.
  • Keep docs/CONTEXT.md as a domain glossary, not a requirements or architecture document.
  • Put decisions in docs/DECISIONS.md; do not hide major tradeoffs inside prose.
  • After architecture is defined, automatically load write-agents-md as a capability skill to generate AGENTS.md. Do not wait for external routing.

Checkpoint Focus

When this node reaches a checkpoint, the Human Check should focus on:

  • whether the chosen technical direction is the simplest viable option
  • module boundaries, shared interfaces, and shared data models
  • storage, external dependency, account, secret, and privacy boundaries
  • decisions that would be expensive to reverse after task breakdown

Inputs

Required:

  • docs/PRD.md

Recommended if present:

  • docs/BRIEF.md
  • docs/NODEFLOW.md
  • docs/ASSUMPTIONS.md
  • existing repository files and config

Actions

  1. Read required inputs.
  2. Identify product constraints that affect architecture:
    • platform
    • data boundary
    • interaction mode
    • storage needs
    • external dependencies
    • validation expectations
  3. Review from four angles:
    • product fit
    • technical feasibility
    • data structure
    • risk
  4. Generate docs/ARCH.md.
  5. Generate docs/CONTEXT.md.
  6. Update docs/DECISIONS.md.
  7. Load write-agents-md as a capability skill and use it to create/update AGENTS.md.
  8. Create CLAUDE.md with only:
@AGENTS.md

ARCH.md Structure

# ARCH

## 技术栈

## 模块边界

## 数据格式

## 文件结构

## 核心依赖

## 风险点

## 共享接口

## 共享数据模型

## 验证方式

CONTEXT.md Rules

docs/CONTEXT.md must contain only project-specific terms and concepts:

# CONTEXT

## 术语

- Term: one-sentence definition.

Rules:

  • one sentence per term
  • no implementation details
  • no requirements
  • no architecture decisions
  • code naming should stay consistent with these terms

DECISIONS.md Rule

Append or update decisions with:

## YYYY-MM-DD - Decision Title

- 决策:
- 理由:
- 影响:
- 替代方案:

Outputs

  • docs/ARCH.md
  • docs/CONTEXT.md
  • docs/DECISIONS.md
  • AGENTS.md
  • CLAUDE.md

Completion Criteria

  • docs/ARCH.md exists and covers tech stack, modules, data, dependencies, risks, interfaces, and shared models.
  • docs/CONTEXT.md exists and contains only glossary content.
  • docs/DECISIONS.md records key architecture choices.
  • AGENTS.md exists and is based on project facts, not guesses.
  • CLAUDE.md exists and points to AGENTS.md.

Stop Conditions

Stop and ask if:

  • docs/PRD.md is missing.
  • the PRD does not define product goal or core requirements clearly enough to choose architecture.
  • tech stack or platform choice is a high-impact decision and cannot be inferred.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.