neat-sdd-analysis

分类: 开发与编程 | 上传者: neatskills | 下载: 0 | 版本: v1.0（最新）
在分析现有代码库时使用 - 在规划和细化之前通过分层分析 (L0-L6) 提取结构化知识
更新日志: Source: GitHub https://github.com/neatskills/neat-sdd
目录结构

当前层级: tree/main/neat_01_analysis/
📄 README.md 2.9 KB
📄 SKILL.md 5.0 KB
SKILL.md

---
name: neat-sdd-analysis
description: Use when analyzing an existing codebase - extracts structured knowledge through layered analysis (L0-L6) before planning and refinement
---

# Analysis

**Role:** You are a software architect who extracts actionable insights from codebases — every finding answers "so what?"

## Overview

Layered knowledge extraction (L0-L6) for unfamiliar codebases.

## When to Use

- Onboarding to codebases
- Pre-refactoring/migration
- Building understanding before changes
- Reference analysis

## Quick Reference

| # | What |
|---|------|
| 1 | Ask: Single/platform? References? |
| 2 | **Platform:** Check → prompt → list/select. **Refs:** Analyze first |
| 3 | Derive name (slugify, max 20) |
| 4 | Check empty → minimal |
| 5 | Check existing, warn |
| 6 | L1-L6, L0 last |
| 7 | Save, update specs.md, offer PDF |
| 8 | Suggest domains/knowledge |

## Setup

1. Ask: Single/platform? References?
2. **Platform:** Check `docs/specs/analysis-platform.md`. **Missing:** Prompt. **Declines:** Warn. **Agrees:** Analyze, list, select.
3. **References:** Ask purpose, analyze FIRST
4. Derive name (slugify, max 20)
5. Check source → empty? Minimal
6. Check existing, warn

---

## Layered Analysis

Sequential L1→L6, L0 last. Do NOT skip. **Stop if fails.**

```mermaid
graph TD
  L1[L1: Foundation] --> L2[L2: Interfaces]
  L2 --> L3[L3: Architecture]
  L3 --> L4[L4: Technical Flows]
  L4 --> L5[L5: Business Logic]
  L5 --> L6[L6: Health]
  L6 --> L0[L0: Executive Summary - write LAST, place FIRST]
```

### L1: Foundation

Tech stack, dependencies, structure (max 30 lines), build. **Platforms:** Context first, then Tech Stack.

### L2: Interfaces

API endpoints, data models, events, config, integrations.

### L3: Architecture

Pattern, component map, deployment, concerns. Flag anti-patterns.

### L4: Technical Flows

3+ flows (user, auth, error), state management, 2-3 edge cases/flow. Mermaid for complex.

### L5: Business Logic

User journeys, workflows, domain concepts, permissions, edge cases.

### L6: Health

Test coverage, tech debt, dead code, security, dependencies. Flag anti-patterns. **Be honest—risks, not praise.**

### L0: Executive Summary (write last, place first)

Purpose, architecture (1 sentence each), stats, top findings, risks, component map.

---

## Platform Analysis

For platforms (multiple products, shared infrastructure).

### Scope

**Analyze:** Workspace, tooling, orchestration, platform services, CI/CD, patterns
**Exclude:** Product code

### Layer Focus

| Layer | Focus |
|-------|-------|
| L1 | Workspace, tools, build, deployment |
| L2 | Platform APIs, config, events |
| L3 | Infrastructure, mesh, architecture |
| L4 | Cross-platform flows |
| L5 | Capabilities, constraints, multi-tenancy |
| L6 | Tech debt, security, dependencies |

**Output:** `analysis-platform.md` at `<repo-root>/docs/specs/`

### Pushback

| Pressure | Response |
|----------|----------|
| "Quickly" | "Prevents rework." |
| "Just [product]" | "Captures deploy, integration." |
| "Standard" | "Custom exists." |
| "Waste time" | "Downstream needs it." |

---

## Empty Handling

Minimal analysis (L0-L6 placeholders, `empty: true`), minimal specs.md.

---

## Output Format

**Filename:** `analysis-<product>.md`, `analysis-<ref-name>.md`, `analysis-platform.md`
**Location:** [output-path.md](../references/output-path.md)
**Frontmatter:** `type`, `analyzed`, `source`/`purpose` (refs only)
**Sections:** `## L[n]: Title` → Findings → Content → Implications

## Process

TodoWrite: ask structure/refs → (platform: check → prompt → list/select) → (refs: first) → derive name → check empty → L1-L6 → L0 → save → update specs.md.

## Common Mistakes

| Mistake | Fix |
|---------|-----|
| File tree dump | Conventions, max 30 lines |
| Happy path only | Add auth, error, edge cases |
| Facts only | Add implications |
| L0 first | Write last |
| Prose | Tables, Mermaid |
| Refs after target | FIRST |
| Missing frontmatter | Add type/analyzed/source/purpose |
| Empty: fail/skip | Minimal + `empty: true` |

## Platform Rationalization

| Thought | Reality |
|---------|---------|
| "User knows" | May not |
| "Just focus" | Can't without |
| "Quickly" | Prevents rework |
| "Standard" | Custom exists |
| "Waste time" | Skipping worse |
| "Docs done" | Docs ≠ L0-L6 |
| "Add later" | Needs now |
| "Don't" | Explain, offer |

**Prompt with reasoning.**

## After Saving

1. **Create/update specs.md**:
   - **Location:** `<repo-root>/docs/specs/specs.md`
   - Sections: Tech Stack, Architecture, Conventions, Avoid, Commands, Knowledge Base
   - **Knowledge Base** ([format](../references/knowledge-registration.md)): `- Analysis: docs/specs/<product>/analysis-<product>.md`
   - **References:** Add Target/References
   - **Platforms:** Add Context, nest Tech Stack
   - **Commands:** Auto-detect
   - Marker: `<!-- Generated by neat-sdd-analysis. Review and customize. -->`
2. **Offer PDF:** "Want PDF? (needs `neat-utils`)" → invoke `neat-util-pdf`
3. **Suggest:** `neat-sdd-knowledge` or `neat-sdd-domains`
登录后下载/点赞/收藏 ❤ 5 | ★ 0
neat-sdd-analysis

目录结构

SKILL.md

举报内容

提示
