How do I turn an AI coding session into a pull request description?
Write the description from three things: what you asked the agent for, the diff, and the test
results. Ask the agent to draft it at the end of the session, then check every line against
git diff. A good one says what changed, why, how it was tested, and what a reviewer
should look at first. In Codex, /review is a second read before you open it. A
BuildGist explanation of the run is a ready first draft of the "what changed" part.
BuildGist is a Mac app for Claude Code, Cursor and Codex. Download for Mac.
What a reviewer needs
## What changed
(one paragraph, plain words)
## Why
(the request or issue, in one or two lines)
## How it was tested
(the tests that ran and their result; anything checked by hand)
## Look at first
(the risky files: sign-in, payments, data, migrations)
## Not in this PR
(what was left for later, and anything the agent tried and dropped)
The last section is the one agent sessions most often lose, and the one reviewers most often ask about.
1. Gather the facts
git diff --stat main...HEAD
git log --oneline main..HEAD
Use your base branch in place of main. Include uncommitted work only if you mean to
commit it. Then run the tests, so "how it was tested" reports a run you saw.
2. Ask the agent for a draft
Before you close the session, while it still knows the reasons:
Write a pull request description for the changes on this branch with these
headings: What changed, Why, How it was tested, Look at first, Not in this PR.
Base it on git diff main...HEAD. Do not claim tests passed unless you ran them
in this session.
This works in Claude Code, Codex and Cursor. The draft reflects what the agent believes it did, so it can describe a change that is not in the diff, or skip one that is.
3. Check it against the diff
Go file by file. Every file in git diff --stat should be accounted for, and nothing
should be described that the diff does not show. Fix the "How it was tested" line to match the
run you just did.
4. Keep the session if reviewers will ask why
- Claude Code:
/export pr-notes.txtsaves the conversation as plain text. - Codex:
/export session.mdsaves it as Markdown.
Do not paste a transcript into the PR. Quote the one decision a reviewer will ask about.
5. Get a second read
Codex's /review and Claude Code's /code-review review the changes
before a person does. Cursor has Agent Review, covered in
how to review Cursor agent changes. A review
from the same kind of tool that wrote the code is useful, but it does not replace reading the
risky files yourself.
If several sessions built the branch
Write the description from the branch, not from the last session. Each session saw only its own part. List the requests in order under "Why", and read the combined diff once.
Where BuildGist fits
After each meaningful run of Claude Code, Cursor or Codex, BuildGist writes a plain-English explanation in its Inbox, uncommitted work included: what changed, and concerns worth checking. It records the request you typed, the files changed with line counts, and the tests that ran with their result, which covers the "Why" and "How it was tested" sections. Copy what fits and edit it. It is written by AI, so check it against the code and the tests like any draft. What an explanation adds over the diff is in BuildGist or the git diff, and the full review steps are in how to review AI code before merging.
Questions
Can Claude Code write my pull request description?
Yes. Ask it to, at the end of the session, from git diff against your base branch. Check each line against the diff and the test run before you use it.
Should a PR description say an AI agent wrote the code?
Follow your team's policy. Either way, the description should say how the change was tested and what a reviewer should check first.
What is the fastest check of an agent's PR draft?
Compare its file list with git diff --stat, and its test claim with a test run you watched.
Sources
- Claude Code commands (/export, /code-review), checked October 25, 2026.
- Codex developer commands (/export, /review), checked October 25, 2026.
What you need
- A Mac with Apple silicon and macOS 12 or later.
- Claude Code, Cursor or Codex.
- A Git project. No commit, push, remote or GitHub repository is needed.
- A GitHub account, to sign in. BuildGist does not access your GitHub repositories.
Codex needs you to approve BuildGist once, in Codex. BuildGist shows the steps. Setup is in Getting started.
To explain a run, your computer sends BuildGist what that run needs, and BuildGist passes it to Anthropic. BuildGist does not keep this material. It does not send your whole repository, files the run did not change, or files named like secrets, such as .env files and keys. It keeps the explanation and short cited excerpts of changed code, at most 12 lines each. Secret masking is best-effort. Privacy has the full list.
Every account gets 7 days of BuildGist Pro free, with no card. The 7 days start at your first explanation or Continue handoff. Nobody is charged when the trial ends. After that, BuildGist Pro is $12 per month, plus any applicable tax shown at checkout. See pricing.
Try BuildGist Download for Mac
Apple silicon, macOS 12 or later. United States, 18 or older.
Related
- What did Claude Code change?
- Switch from Claude Code to Codex
- Understand code an agent wrote
- Hand off Claude Code to Cursor
- Review what Codex changed
- You’ve hit your session limit in Claude Code
- You’ve hit your usage limit in Codex
- Keep context across agents
- Review AI code before merging
- BuildGist or the git diff
- Claude Code session history
- Undo Claude Code changes
- BuildGist vs CLAUDE.md, AGENTS.md and HANDOFF.md
- Keep context from Claude Code to Cursor
- Resume an agent session in another tool
- BuildGist vs Conductor
- Codex context window full
- Undo Codex changes
- Cursor usage limit reached
- Review Cursor agent changes
- Codex session history location
- Your conversation is too long in Cursor
- One default coding agent, or two?
- Tools that summarize agent sessions
- Switch from Codex to Claude Code
- Move from Cursor to Claude Code
- Switch from GitHub Copilot to Claude Code
- Move from Windsurf to Cursor
- Switch from Gemini CLI to Claude Code
- Coding agent crashed mid-task
- Claude Code lost context after compacting
- What commands did the agent run?
- Which coding agents read AGENTS.md?
Questions: kayvanandre@gmail.com.