Skip to content

Commit e0d518f

Browse files
wimi321wimi321
committed
Tighten public-facing copy
1 parent ac478fb commit e0d518f

6 files changed

Lines changed: 64 additions & 78 deletions

File tree

‎README.md‎

Lines changed: 19 additions & 25 deletions
Original file line numberDiff line numberDiff line change
@@ -6,42 +6,37 @@
66
<img src="./assets/hero-banner.svg" alt="Task Bundle hero banner" width="100%" />
77
</p>
88

9-
<p align="center"><strong>Turn AI coding runs into portable, replayable, benchmark-ready task bundles.</strong></p>
10-
<p align="center">A practical format between raw chat logs and heavyweight benchmark platforms.</p>
9+
<p align="center"><strong>Turn AI coding runs into portable, replayable task bundles.</strong></p>
10+
<p align="center">Useful when chat logs are too loose and full benchmark platforms are too heavy.</p>
1111
<p align="center">
1212
<a href="#quickstart"><strong>Quick Start</strong></a> ·
13-
<a href="#real-bundles"><strong>Real Output</strong></a> ·
14-
<a href="#format-vs-alternatives"><strong>Why This Format</strong></a> ·
13+
<a href="#example-output"><strong>Example Output</strong></a> ·
14+
<a href="#where-it-fits"><strong>Where It Fits</strong></a> ·
1515
<a href="./docs/bundle-format.md"><strong>Bundle Format</strong></a> ·
1616
<a href="./docs/sample-benchmark-report.md"><strong>Sample Report</strong></a> ·
17-
<a href="./ROADMAP.md"><strong>Roadmap</strong></a> ·
18-
<a href="./docs/branding.md"><strong>Brand Assets</strong></a>
17+
<a href="./ROADMAP.md"><strong>Roadmap</strong></a>
1918
</p>
2019

2120
[![CI](https://github.com/wimi321/task-bundle/actions/workflows/ci.yml/badge.svg)](https://github.com/wimi321/task-bundle/actions/workflows/ci.yml)
2221
[![GitHub stars](https://img.shields.io/github/stars/wimi321/task-bundle?style=social)](https://github.com/wimi321/task-bundle/stargazers)
2322
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](./LICENSE)
2423

25-
Task Bundle is a TypeScript + Node.js CLI for teams building agents, evals, coding benchmarks, and reproducible AI workflows.
24+
Task Bundle is a TypeScript + Node.js CLI for packaging one coding task into a directory you can inspect, compare, archive, validate, and report on.
2625

27-
Package a task once, inspect it later, compare tools on the same starting point, and generate benchmark-style reports from real artifacts.
26+
Use it to:
27+
- keep task inputs, summaries, diffs, events, and workspace files together
28+
- compare Codex, Claude Code, Cursor, or internal tools using metadata, hashes, and outcome fields
29+
- generate benchmark-style reports from a folder of bundles
30+
- preserve enough context for reruns without aiming for token-perfect replay
2831

29-
It helps you:
30-
- turn one AI coding run into a clean, shareable directory instead of leaving it scattered across screenshots, transcripts, or loose patches
31-
- compare Codex, Claude Code, Cursor, or internal agents using metadata, hashes, and outcome fields
32-
- generate benchmark-style reports from a folder of bundles without standing up a full evaluation platform first
33-
- preserve enough context for reruns and comparisons without requiring token-perfect recording
34-
35-
It fits the gap between raw logs and full evaluation systems: light enough for day-to-day work, structured enough for replay and benchmarking.
36-
37-
It is designed for workflows where you want to:
32+
It works best when you want to:
3833
- inspect what happened
3934
- share a task with someone else
4035
- rerun a task later
4136
- compare outputs across tools and models
4237
- grow toward replay and benchmark workflows
4338

44-
It is intentionally not:
39+
It is not:
4540
- an agent framework
4641
- a chat UI
4742
- a provider router
@@ -60,13 +55,13 @@ npm run build
6055
npm run dev -- compare ./examples/hello-world-bundle ./examples/hello-world-bundle-claude
6156
```
6257

63-
If you want the shortest possible proof that the project already works, this is it.
58+
This is the fastest way to see the format in action.
6459

6560
![Task Bundle workflow overview](./assets/workflow-overview.svg)
6661

67-
<a id="real-bundles"></a>
62+
<a id="example-output"></a>
6863

69-
## See It On Real Bundles
64+
## Example Output
7065

7166
Inspect a bundle:
7267

@@ -110,13 +105,13 @@ Ranking
110105
2. Fix greeting punctuation | claude-code / claude-sonnet-4 | success | score 0.89
111106
```
112107

113-
Browse the committed example report:
108+
See the committed sample report:
114109
- [docs/sample-benchmark-report.md](./docs/sample-benchmark-report.md)
115110
- [docs/sample-benchmark-report.zh-CN.md](./docs/sample-benchmark-report.zh-CN.md)
116111

117-
<a id="format-vs-alternatives"></a>
112+
<a id="where-it-fits"></a>
118113

119-
## How It Compares To Common Alternatives
114+
## Where It Fits
120115

121116
| Need | Chat logs | Zip or tarball | Full benchmark platform | Task Bundle |
122117
| --- | --- | --- | --- | --- |
@@ -161,7 +156,6 @@ See:
161156
- [docs/bundle-format.zh-CN.md](./docs/bundle-format.zh-CN.md)
162157
- [docs/design-decisions.md](./docs/design-decisions.md)
163158
- [docs/replay-contract.md](./docs/replay-contract.md)
164-
- [docs/branding.md](./docs/branding.md)
165159

166160
## Five-Minute Demo
167161

‎README.zh-CN.md‎

Lines changed: 24 additions & 30 deletions
Original file line numberDiff line numberDiff line change
@@ -6,42 +6,37 @@
66
<img src="./assets/hero-banner.svg" alt="Task Bundle hero banner" width="100%" />
77
</p>
88

9-
<p align="center"><strong>把 AI coding 过程变成可分享、可重跑、可比较、可做 benchmark 的任务包。</strong></p>
10-
<p align="center">它适合放在聊天记录和 benchmark 平台之间，承接真实任务与结果。</p>
9+
<p align="center"><strong>把 AI coding 过程整理成可分享、可比较、可重跑的任务包。</strong></p>
10+
<p align="center">适合用在聊天记录不够稳定、完整 benchmark 平台又太重的场景里。</p>
1111
<p align="center">
1212
<a href="#quickstart"><strong>快速开始</strong></a> ·
13-
<a href="#real-bundles"><strong>真实输出</strong></a> ·
14-
<a href="#format-vs-alternatives"><strong>为什么是这个格式</strong></a> ·
13+
<a href="#example-output"><strong>示例输出</strong></a> ·
14+
<a href="#where-it-fits"><strong>方案对比</strong></a> ·
1515
<a href="./docs/bundle-format.zh-CN.md"><strong>格式说明</strong></a> ·
1616
<a href="./docs/sample-benchmark-report.zh-CN.md"><strong>示例报告</strong></a> ·
17-
<a href="./ROADMAP.zh-CN.md"><strong>路线图</strong></a> ·
18-
<a href="./docs/branding.zh-CN.md"><strong>品牌素材</strong></a>
17+
<a href="./ROADMAP.zh-CN.md"><strong>路线图</strong></a>
1918
</p>
2019

2120
[![CI](https://github.com/wimi321/task-bundle/actions/workflows/ci.yml/badge.svg)](https://github.com/wimi321/task-bundle/actions/workflows/ci.yml)
2221
[![GitHub stars](https://img.shields.io/github/stars/wimi321/task-bundle?style=social)](https://github.com/wimi321/task-bundle/stargazers)
2322
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](./LICENSE)
2423

25-
Task Bundle 是一个 TypeScript + Node.js CLI，适合 agent、eval、benchmark、可复现实验这类工作流。
24+
Task Bundle 是一个 TypeScript + Node.js CLI，用来把一次编码任务打包成可以查看、比较、归档、校验、生成报告的目录。
2625

27-
把一次运行整理好之后，就可以 inspect、compare、validate、report，也方便把不同工具放到同一起点上做对照。
26+
你可以用它来：
27+
- 把任务输入、执行摘要、diff、事件和工作区文件放在一起
28+
- 比较 Codex、Claude Code、Cursor 或内部工具的运行结果，并保留元数据、哈希和 outcome 字段
29+
- 从一组 bundle 生成 benchmark 风格报告
30+
- 为后续重跑保留足够上下文，而不是追求逐 token 回放
2831

29-
它主要解决这些问题：
30-
- 把一次 AI coding 任务整理成干净、稳定、可搬运的目录，而不是散落在截图、聊天记录或 patch 里
31-
- 比较 Codex、Claude Code、Cursor 或内部工具的结果，而且比较依据包括元数据、哈希和 outcome 字段
32-
- 从一组 bundle 直接生成 benchmark 风格报告，不用先搭完整评测平台
33-
- 为后续重跑和比较保留足够上下文，而不是依赖逐 token 录制
34-
35-
它适合放在“聊天记录不够稳”和“完整 benchmark 平台太重”之间，作为更轻但足够结构化的方案。
36-
37-
它适合这些场景：
32+
它比较适合这些场景：
3833
- 查看一次任务最后到底做了什么
3934
- 把任务结果分享给别人
4035
- 在之后重新执行同一个任务
4136
- 比较不同模型或工具在同一起点上的表现
4237
- 作为未来 replay / benchmark 工作流的基础层
4338

44-
它明确不做这些事情：
39+
它不打算解决这些问题：
4540
- agent 框架
4641
- 聊天 UI
4742
- provider 路由器
@@ -60,15 +55,15 @@ npm run build
6055
npm run dev -- compare ./examples/hello-world-bundle ./examples/hello-world-bundle-claude
6156
```
6257

63-
如果你只想先确认“这项目现在到底能不能用”，这组命令就是最短路径。
58+
这是最快看懂项目在做什么的一组命令。
6459

6560
![Task Bundle workflow overview](./assets/workflow-overview.svg)
6661

67-
<a id="real-bundles"></a>
62+
<a id="example-output"></a>
6863

69-
## 看看真实输出
64+
## 示例输出
7065

71-
先 inspect 一个 bundle：
66+
先查看一个 bundle：
7267

7368
```text
7469
$ npm run dev -- inspect ./examples/hello-world-bundle
@@ -110,31 +105,31 @@ Ranking
110105
2. Fix greeting punctuation | claude-code / claude-sonnet-4 | success | score 0.89
111106
```
112107

113-
你也可以直接点开仓库里提交好的示例报告：
108+
也可以直接查看仓库里提交好的示例报告：
114109
- [docs/sample-benchmark-report.zh-CN.md](./docs/sample-benchmark-report.zh-CN.md)
115110
- [docs/sample-benchmark-report.md](./docs/sample-benchmark-report.md)
116111

117-
<a id="format-vs-alternatives"></a>
112+
<a id="where-it-fits"></a>
118113

119-
## 和常见替代方案怎么区分
114+
## 和常见方案对比
120115

121116
| 需求 | 聊天记录 | Zip / tarball | 完整 benchmark 平台 | Task Bundle |
122117
| --- | --- | --- | --- | --- |
123118
| 把原始任务和最终结果放在一起分享 | 部分满足 | 可以 | 可以 | 可以 |
124-
| 在同一起点上比较不同工具 | 很弱 | 很靠手工 | 可以 | 可以 |
119+
| 在同一起点上比较不同工具 | 较弱 | 很靠手工 | 可以 | 可以 |
125120
| 携带 artifact 哈希和结果元数据 | 不行 | 不行 | 可以 | 可以 |
126121
| 足够轻，能融入日常 coding 工作流 | 可以 | 可以 | 不太行 | 可以 |
127-
| 之后继续长成 replay / benchmark 工作流 | 很弱 | 很弱 | 可以 | 可以 |
122+
| 之后继续扩展成 replay / benchmark 工作流 | 较弱 | 较弱 | 可以 | 可以 |
128123

129124
## 为什么值得关注
130125

131126
很多 AI coding 结果最后只留下截图、聊天记录或者一个 patch，后续几乎没法稳定比较。
132127

133-
Task Bundle 想解决的就是这个空档：把一次任务变成一个可以 inspect、archive、compare、validate、report 的稳定单元。它特别适合：
128+
Task Bundle 主要解决的是这个问题：把一次任务变成一个可以查看、归档、比较、校验、生成报告的稳定单元。它比较适合：
134129
- 想做可复现实验的 agent 作者
135130
- 想做任务评测和 benchmark 的团队
136131
- 想比较 Codex、Claude Code、Cursor 或内部工具的开发者
137-
- 更关心可重跑，而不是逐 token 回放一致性的人
132+
- 更关心可重跑，而不是逐 token 一致性的人
138133

139134
## 这里的 Replay 是什么意思
140135

@@ -161,7 +156,6 @@ task-bundle/
161156
- [docs/bundle-format.md](./docs/bundle-format.md)
162157
- [docs/design-decisions.md](./docs/design-decisions.md)
163158
- [docs/replay-contract.md](./docs/replay-contract.md)
164-
- [docs/branding.zh-CN.md](./docs/branding.zh-CN.md)
165159

166160
## 五分钟演示
167161

‎docs/branding.md‎

Lines changed: 5 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1,19 +1,17 @@
11
# Branding Assets
22

3-
Task Bundle includes repository-ready visual assets under `assets/`.
4-
5-
The art direction is intentionally warm-editorial rather than generic SaaS gradients: a calm dark field, paper-toned bundle cards, and a benchmark signal accent that reinforces "portable tasks" plus "measurable outcomes."
3+
This repository includes the visual assets used in the README and GitHub social preview.
64

75
## Files
86

97
- `assets/hero-banner.svg`
10-
Embedded at the top of the README to make the repository landing page feel like a product, not just a package listing.
8+
Hero image used at the top of the README.
119
- `assets/workflow-overview.svg`
12-
A second README visual that explains the capture -> inspect -> compare -> report loop in one glance.
10+
Workflow diagram used in the README.
1311
- `assets/social-preview.svg`
1412
Source artwork for GitHub social preview uploads.
1513
- `assets/social-preview.png`
16-
Recommended raster export for GitHub social preview uploads. Kept in the repository for easy manual upload, but not required in the npm package.
14+
Recommended raster export for GitHub social preview uploads.
1715

1816
## Suggested GitHub Setup
1917

@@ -23,4 +21,4 @@ The art direction is intentionally warm-editorial rather than generic SaaS gradi
2321

2422
## Local Export Tips
2523

26-
If you want to regenerate the PNG on macOS, you can use Quick Look or another SVG-to-PNG tool. The repository artwork is intentionally kept as SVG so it stays editable and versionable, while the committed PNG keeps GitHub social preview setup friction low.
24+
The SVG files stay in the repository so they remain editable and easy to version.

‎docs/branding.zh-CN.md‎

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1,17 +1,17 @@
11
# 品牌素材
22

3-
Task Bundle 在 `assets/` 目录下提供了一套可直接用于仓库展示的视觉素材，让 README、GitHub 首页和分享卡片能保持统一气质。
3+
这个仓库在 `assets/` 目录下提供了 README 和 GitHub 社交预览图会用到的视觉素材。
44

55
## 文件说明
66

77
- `assets/hero-banner.svg`
8-
中英文 README 顶部使用的主视觉横幅，可继续编辑。
8+
README 顶部使用的主视觉横幅。
99
- `assets/workflow-overview.svg`
10-
README 里的第二张主视觉，用来一眼解释 capture -> inspect -> compare -> report 这条路径。
10+
README 中使用的工作流示意图。
1111
- `assets/social-preview.svg`
1212
GitHub 社交预览图的可编辑源文件。
1313
- `assets/social-preview.png`
14-
已导出的上传版本，适合直接放到 GitHub 仓库设置里。
14+
GitHub 社交预览图的 PNG 版本。
1515

1616
## 推荐设置
1717

@@ -27,4 +27,4 @@ Task Bundle 在 `assets/` 目录下提供了一套可直接用于仓库展示的
2727
sips -s format png ./assets/social-preview.svg --out ./assets/social-preview.png
2828
```
2929

30-
仓库里保留 SVG，是为了让这套素材更容易继续修改、做版本对比，也更适合长期维护。
30+
仓库里保留 SVG，是为了继续编辑和版本管理更方便。

‎docs/sample-benchmark-report.md‎

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
# Sample Benchmark Report
22

3-
This page shows what `taskbundle report` looks like against the example bundles included in this repository.
3+
This page keeps a saved example of `taskbundle report` generated from the example bundles in this repository.
44

55
## Regenerate Locally
66

@@ -28,8 +28,8 @@ npm run dev -- report ./examples --out ./dist/benchmark-report.md
2828
| codex | gpt-5 | 1 | 1 | 1 | 0.93 | 0.93 |
2929
| claude-code | claude-sonnet-4 | 1 | 1 | 1 | 0.89 | 0.89 |
3030

31-
## Why This Matters
31+
## Why Keep This Page
3232

33-
- It gives the repo a benchmark-shaped artifact without forcing a full benchmark platform.
34-
- It shows that the example bundles are not toy files with no downstream use.
35-
- It makes cross-tool comparisons legible for humans before you build dashboards.
33+
- It gives readers a concrete example of the report output.
34+
- It shows how the example bundles can be compared without extra tooling.
35+
- It provides a stable link that can be referenced from the README.

‎docs/sample-benchmark-report.zh-CN.md‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
# 示例 Benchmark 报告
22

3-
这个页面展示的是：把仓库自带的 example bundles 交给 `taskbundle report` 之后，大概会得到什么样的结果。
3+
这个页面保存了一份基于仓库示例 bundle 生成的 `taskbundle report` 输出，方便直接查看结果长什么样。
44

55
## 本地重新生成
66

@@ -21,15 +21,15 @@ npm run dev -- report ./examples --out ./dist/benchmark-report.md
2121
| 1 | Fix greeting punctuation | codex | gpt-5 | success | 0.93 | 3 | 1 |
2222
| 2 | Fix greeting punctuation | claude-code | claude-sonnet-4 | success | 0.89 | 4 | 1 |
2323

24-
## Tool / Model 排行
24+
## 按工具 / 模型汇总
2525

2626
| Tool | Model | Runs | Scored | Successes | Avg Score | Best Score |
2727
| --- | --- | --- | --- | --- | --- | --- |
2828
| codex | gpt-5 | 1 | 1 | 1 | 0.93 | 0.93 |
2929
| claude-code | claude-sonnet-4 | 1 | 1 | 1 | 0.89 | 0.89 |
3030

31-
## 为什么这个页面有价值
31+
## 为什么保留这个页面
3232

33-
- 它让仓库直接具备一个 benchmark 风格的可见成果，不需要先做完整平台。
34-
- 它说明 example bundles 不是摆设，而是真的可以继续拿来分析和比较。
35-
- 它让“跨工具比较”在没有 dashboard 之前，也已经足够清楚可读。
33+
- 让读者直接看到报告输出的样子。
34+
- 说明仓库里的示例 bundle 可以继续拿来比较和分析。
35+
- README 可以稳定链接到这份示例结果。

0 commit comments

Comments
 (0)