用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/mlflow/dev --skill fetch-diff命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | fetch-diff |
| description | Fetch PR diff with filtering and line numbers for code review. |
Fetches a pull request diff and adds line numbers for easier review comment placement. Auto-generated files are shown with masked diffs.
uv run --package skills skills fetch-diff <pr_url> [--files <pattern> ...]
Examples:
# Fetch the full diff
uv run --package skills skills fetch-diff https://github.com/mlflow/mlflow/pull/123
# Fetch only Python files
uv run --package skills skills fetch-diff https://github.com/mlflow/mlflow/pull/123 --files '*.py'
# Fetch only frontend files
uv run --package skills skills fetch-diff https://github.com/mlflow/mlflow/pull/123 --files 'mlflow/server/js/*'
# Multiple patterns
uv run --package skills skills fetch-diff https://github.com/mlflow/mlflow/pull/123 --files '*.py' '*.ts'
Token is auto-detected from GH_TOKEN env var or gh auth token.
Regular file:
diff --git a/path/to/file.py b/path/to/file.py
index abc123..def456 100644
--- a/path/to/file.py
+++ b/path/to/file.py
@@ -10,7 +10,7 @@
10 10 | import os
11 11 | import sys
12 12 | from typing import Optional
13 | -from old_module import OldClass
14 | +from new_module import NewClass
14 15 |
15 16 | def process_data(input_file: str) -> dict:
Auto-generated file (masked):
diff --git a/uv.lock b/uv.lock
index abc123..def456 100644
--- a/uv.lock
+++ b/uv.lock
[Auto-generated file - diff masked]
Deleted file (masked):
diff --git a/path/to/removed.py b/dev/null
index abc123..0000000 100644
--- a/path/to/removed.py
+++ /dev/null
[Deleted file - diff masked]
Each line is annotated as old_line new_line | <marker> content:
- marker (left number only) -> deleted line, side=LEFT, line=old_line+ marker (right number only) -> added line, side=RIGHT, line=new_lineside=RIGHT, line=new_lineReview a pull request and emit a validated review payload.
Add a review comment to a GitHub pull request.
Review a GitHub PR's UI/UX changes by launching the MLflow web app, driving a headless agent-browser over the changed surfaces, and writing a Markdown UI-review comment body (findings + screenshots) for the workflow to post.