mager-bench

Llama 3.3 70B × readme-writer

6.0

Pure documentation ability — no code at all

correctness 6.0quality 6.0documentation 6.01290ms
$ cat readme-writer.prompt — what the model was asked
Write a README.md for a command-line tool called `snapdiff` that compares two directories and reports added, removed, and changed files. The tool is written in Python, installable via pip. Include: a one-line description, installation instructions, usage examples with flags (--ignore-hidden, --output json|text, --depth N), output format explanation, and a short 'How it works' section. Use proper Markdown formatting.

run 1 of 1

6.0
correctness 6.0quality 6.0documentation 6.01290ms

# judge: Covers all required sections and flags but output format section lacks concrete example output, missing sample JSON/text output blocks and a title-only H1 without top-level structure like a Table of Contents or License section.

$ cat llama-3.3-70b.readme-writer.r1.response
# Snapdiff
**A command-line tool to compare two directories and report added, removed, and changed files.**

## Installation
To install `snapdiff`, run the following command:
```bash
pip install snapdiff
```
This will install the `snapdiff` package and make the `snapdiff` command available in your terminal.

## Usage
The basic usage of `snapdiff` is:
```bash
snapdiff dir1 dir2
```
This will compare the contents of `dir1` and `dir2` and report any added, removed, or changed files.

You can customize the output with the following flags:

* `--ignore-hidden`: Ignore hidden files and directories (those that start with a dot `.`)
* `--output`: Specify the output format, either `json` or `text` (default)
* `--depth`: Specify the maximum depth to compare, e.g. `--depth 2` will only compare the top two levels of the directory tree

Examples:
```bash
snapdiff --ignore-hidden dir1 dir2
snapdiff --output json dir1 dir2
snapdiff --depth 2 dir1 dir2
```
You can combine multiple flags to customize the output:
```bash
snapdiff --ignore-hidden --output json --depth 2 dir1 dir2
```
## Output Format
The default output format is a human-readable text format, which shows the added, removed, and changed files in a concise manner. The JSON output format is more verbose and includes additional metadata, such as file timestamps and sizes.

## How it works
`snapdiff` works by recursively traversing the directory trees of `dir1` and `dir2`, comparing the contents of each directory and file. It uses a combination of file metadata (e.g. timestamps, sizes) and file contents to determine whether a file has been added, removed, or changed. The results are then reported in the specified output format.