tt-a1i

Impl Explain — Git skill for Claude Code

Git community

Cross-agent (Claude Code / Codex / opencode) skill that renders implementation plan + git diff into a single-file HTML report — TL;DR, architecture diagram, decision cards, risk snapshot.

How to install Impl Explain

This entry records only its repository, not the path inside it, so there is no exact command to give. Open tt-a1i/impl-explain and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Impl Explain does

Cross-agent (Claude Code / Codex / opencode) skill that renders implementation plan + git diff into a single-file HTML report — TL;DR, architecture diagram, decision cards, risk snapshot. Editorial visual style, no pip deps.

Alternatives in Git

  • Triage Open Pull Requests — Review, label, and act on all open PRs for a repository using parallel review agents 10.8k ★
  • Codebase To Course — by Zara Zhang - A Claude Code skill that turns any codebase into a beautiful, interactive single-page HTML cou 2.1k ★
  • Droidconkotlin — by touchlab - Delivers comprehensive Gradle commands for cross-platform Kotlin Multiplatform development with 1.1k ★

README

**中文** · [English](README.en.md)

impl-explain

**Cross-agent skill for one-page HTML implementation reports**

把"刚做完的实施 + git 历史"渲染成一份**单文件 HTML 叙事报告**。 开会前 30 秒发到 Slack,团队读 60 秒对齐"做了什么 / 为什么这么做 / 有什么风险"。


[![Agents](https://img.shields.io/badge/Agents-Claude_Code_·_Codex_·_opencode-b04a1f)](#compatibility) [![Prompt-only](https://img.shields.io/badge/Prompt--only-no_code_runtime-4a6a2c)](#how-it-works) [![Status](https://img.shields.io/badge/Status-v2_lite-555)](#roadmap)

[Install](#install) · [Usage](#usage) · [What you get](#what-you-get) · [Why](#why-this-exists)


![preview](docs/hero.png)


Why this exists

AI agent 写代码极快——快到团队成员根本来不及看清实施方案、架构、决策。

代码细节随时 `git diff` 拿得到。但**只存在于实施当时**的信息呢?

  • 你为什么选 A 不选 B
  • 哪些事故意没做
  • 哪些事可能出问题

`impl-explain` 把这些信息从 plan + git context 里抽出来,渲染成一份单文件 HTML—— **不是 changelog,不是 dashboard,是叙事。**

How it works

**Prompt-only skill — 没有 Python 脚本、没有 JSON schema、没有 validator**。

Agent 读 `SKILL.md`(指令)+ `examples/sample.html`(视觉参考),然后直接生成 HTML 写到磁盘。一致性靠 sample.html 当 anchor,质量靠 SKILL.md 里的检查清单。

历史版本:v1 走 JSON schema + `render.py` 渲染管线,迭代到第 7 轮后认定"过度设计"。v1 标签:`v1-structured-2026-05-11`,旧评估报告在 [`research/archive/`](research/archive/)。

What you get

每份 HTML 报告大致包含(缺什么跳什么,**不硬凑**):

Section 内容
Hero 衬线大标题 · 副标题 · 1-3 个 metric chip · plan path · commits 折叠
TL;DR 三行:做什么 / 怎么做 / 代价(整体账)+ 尾部风险预告 + → Risks 锚
Architecture Mermaid 流程图 · 可选上方 summary 叙述 · 下方 caption 颜色编码说明
Before / After 数据流改动前后并排对比(mermaid)
Decisions 编号决策卡片:结论式短标题 · 采用 / 理由 / 放弃 / 代价 · chosendeferred
Risks 风险列表 · 顶部 snapshot · Top risk inline highlight · severity 染色 · mitigation 三态
Out of Scope 故意没做的事
顶部 sticky TOC(6 锚点)+ 滚动 progress bar

视觉风格:**浅色编辑风**(warm cream + Fraunces 衬线 + Inter 正文 + JetBrains Mono 标签),不是 dashboard。

完整样例 → [`examples/sample.html`](examples/sample.html)(git clone 后直接双击)。