Skip to main content
skill CX Codex

pr-description · minimax-ai — Claude Code skill: what it does, how to install

pr-description is a Claude Code skill that drafts and verifies pull request descriptions for GitHub and GitLab. It runs when opening a PR or when MCode Review reports missing sections.

MIT ★ 741 updated 2026-11-11

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add MiniMax-AI/mcode-stack
/plugin install jiachen@mcode-stack
Open on GitHub

What pr-description does

pr-description generates pull request summaries or audits existing text against git diffs. Its layout covers context, architectural changes, verification results, rollback steps, and open items. When .github/pull_request_template.md or GitLab templates exist, the skill keeps their headers and inserts its sections into them.

During verification, the skill compares the description against actual code changes to flag unlisted behavior modifications. It prints the draft before publishing. Sending text upstream requires GitHub CLI gh or GitLab CLI glab, and the tool never edits titles, labels, or target branches on its own.

What it helps with

  • draft structured pull request descriptions with rollback steps
  • audit PR text against actual git diff changes
  • fill repository pull request templates without dropping sections
  • publish descriptions directly via GitHub CLI or GitLab CLI

Example prompts for the agent

  • Draft a PR description for the current branch changes
  • Review the PR description against git diff and flag omissions
  • Update the merge request text via glab with rollback details

How to install pr-description

1Plugin marketplace

/plugin marketplace add MiniMax-AI/mcode-stack
/plugin install jiachen@mcode-stack

This installs the whole jiachen plugin with all its skills

2Skills CLI (Vercel)

npx skills add MiniMax-AI/mcode-stack --skill pr-description -a claude-code

3Manually: put the files in place

~/.claude/skills/pr-description/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/pr-description/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about pr-description

How to install pr-description in Claude Code?

Run /plugin marketplace add MiniMax-AI/mcode-stack and then run /plugin install jiachen@mcode-stack.

What does pr-description do?

It drafts structured pull request descriptions, compares descriptions against git diffs, and highlights missing verification or rollback details.

What is required to run pr-description?

Drafting text requires only the agent. Automatic submission to GitHub or GitLab requires the gh or glab CLI tool installed locally.

Source text

SKILL.md · MIT
---
name: pr-description
description: Draft or check a pull request description in a fixed format (background, changes, behavior changes, out of scope, verification, rollback, not done, related) for GitHub or GitLab. Use when opening a PR, when the MCode Review summary reports missing description sections, or when the user asks for a PR description.
---

# PR description

On GitLab a PR is a merge request; the format is the same on both hosts.

Write the description in the user's language, or in the output language a prompt names. A PR opened through a CLI or an API does not get the repository's PR template, so write every section yourself. When the repository has a template (`.github/pull_request_template.md` on GitHub, `.gitlab/merge_request_templates/` on GitLab), keep its headings and checklists and fit the sections below into it.

## When it is required

MCode Review asks for the verification and rollback sections on L and XL PRs and on PRs that touch risk paths. Write all sections for every PR anyway; they cost little and answer the reviewer's first questions.

## Template

```markdown
## Background
<the problem, who hits it, how it shows up; link the ticket or discussion>

## Changes
<what changed, by mechanism, not by file list>

Behavior changes:
- <each user-visible or caller-visible change; write "None" when there is none>

## Out of scope
- <things a reviewer might expect but this PR leaves alone, and why>

## Verification
- <the test that failed before the fix and passes after it, with its name>
- <commands run and their result>

## Rollback
<how to revert, and what happens to data, config or clients written by the new version>

## Not done
- <checks not done, such as manual testing on a platform>

## Related
- <related PRs and how they interact>
```

In Chinese, keep the same sections in the same order with these headings: 背景, 改动 (with 行为变化:, and 无 when there is none), 明确不改的, 验证, 回滚方式, 未执行, 关联.

## Rules

1. Every claim must be checkable. Write "behavior unchanged" (行为不变) only after checking it, and say how.
2. List deployment prerequisites (DDL, config center changes, publish steps) under Changes and name who runs them.
3. Put UI changes' before/after screenshots or recordings under Verification.
4. Keep Not done honest; the reviewer treats missing checks as open questions, not as failures.

## Checking a description

Read the PR's diff as well as its description, then report each missing or empty section, each claim the diff does not support, each behavior change in the diff that the description leaves out, and a missing rollback plan for changes to stored data or config. Suggest the missing text; leave sections that are fine alone.

## Posting

Show the text first, and post it only when the user asks. Write it to a file, then:

- GitHub: `gh pr create --title "<title>" --body-file <file>` for a new PR, `gh pr edit <n> --body-file <file>` for an existing one.
- GitLab: `glab mr create --title "<title>" --description-file <file>`, or `glab mr update <n> --description-file <file>`. Older `glab` versions without `--description-file` take `--description "$(cat <file>)"`.

Do not change a PR's title, labels, reviewers or target branch unless the user asks.

© Jiachen (MiniMax Code team) · MIT · MiniMax-AI/mcode-stack

Similar skills

All skills for Git and GitHub →

pr-description by other authors comes first.

From the same set27

The whole set →
harness-experiment Harness-experiment is a Claude Code skill that runs controlled experiments to validate agent scaffolding changes. It triggers when adjusting prompts, tools, or subagent layouts for performance tracking. CX Codex +1 ★ 741repo stars independent-review independent-review is a Claude Code skill that reviews changes from a clean context before merge. It compares the diff with the requirement and test output. CX Codex +1 ★ 741repo stars agent-brief The agent-brief skill for Claude Code structures task delegation to AI agents. It forces clarity on stakes, hypotheses, and verification criteria before an agent starts working. CX Codex +1 ★ 741repo stars evidence-first-review evidence-first-review is a Claude Code skill that audits an agent’s finished work from command logs and test evidence. It runs before a PR merge or an analysis is accepted. CX Codex +1 ★ 741repo stars subagent-split-check subagent-split-check is a Claude Code skill that decides whether work belongs in a subagent or the main loop. It checks context boundaries before a split. CX Codex +1 ★ 741repo stars context-tax-audit The context-tax-audit skill for Claude Code inventories injected data in agent requests to reduce token usage. It identifies redundant prompt blocks and configurations that hinder prompt cache performance. CX Codex +1 ★ 741repo stars long-session-check long-session-check is a Claude Code skill that audits long-session context and memory handling. It starts when constraints get lost or compaction has happened several times. CX Codex +1 ★ 741repo stars long-command-check long-command-check is a Claude Code skill that audits shell command execution in agent harnesses. It checks dev server hangs, timeouts, backgrounding, and prompt cache expiration. CX Codex +1 ★ 741repo stars

Project sponsors

Companies that keep this analytics open