AgentSkillsCN

local-review-comments

从review-comments.json文件中提取本地评审意见,并将其格式化为Markdown,供LLM读取与理解。适用于用户希望处理本地评审意见或反馈时的场景。

SKILL.md
--- frontmatter
name: local-review-comments
description: Retrieves local review comments from the review-comments.json file and formats them as markdown for LLM consumption. Use when the user asks to address local review comments or feedback.

Retrieving local review comments

Use the repo-to-md CLI tool to format local review comments as LLM-friendly markdown.

Primary usage

bash
repo-to-md local format

This reads review-comments.json from the current directory and outputs markdown to stdout.

Prerequisites

  • A review-comments.json file must exist (created by repo-to-md local review)
  • The file contains comments from a local review session

Alternative usage

Specify a different comments file:

bash
repo-to-md local format path/to/comments.json

Output to a file instead of stdout:

bash
repo-to-md local format -o output.md

Output format

Generates markdown with file sections containing code context and embedded review comments:

markdown
# Review comments

Please address the following review comments. The comments are given as inline
comments inside code blocks. They use the comment format of the language of the
commented file. Each comment is wrapped in "<review>" tags.

For example, a Rust file may look like:

```rust
fn main() {
    println!("13");
    // <review>
    // Please use 42 as the number
    // </review>
}
```

## General Comments

**@user:**

This is a general comment about the overall changes.

## `src/main.rs` - Line 10

```rust
pub fn main() {
// <review>
// Consider using clap for argument parsing
// </review>
    println!("Hello");
}
```

Each comment is embedded as <review>...</review> XML tags using the language-specific comment syntax. Large diffs are intelligently truncated to show only context around commented lines.