Retype the changes an agent made to your feature
They inspect the key edits from an agent-built feature and type them in themselves to better understand the code.
- What they did
- After a custom-built agent produced a new feature, fcoury used a skill that duplicated the starting branch and surfaced the changes that mattered most, while leaving out imports and other boilerplate. They then entered each important edit manually in an editor so they could learn the implementation more deeply. They also experimented with asking targeted questions about particular diffs and with prompts that left gaps to fill.
- How it went
- This approach was useful when the work was critical, but it was too slow to become a regular habit. The question-and-gap versions helped some, and they are still searching for better ways and more research.
- Tips
- Use this for features where understanding the code matters more than speed. For smaller changes, the extra time probably outweighs the benefit.
Try it yourself with Claude Code
I just had you build [feature name] on [branch name]. Please compare that work with a fresh copy of [base branch] and send me only the diffs that matter most, one chunk at a time. Leave out imports and other routine changes, and add a brief note on why each chunk is important. Don’t edit any files; I’ll type each chunk myself. Wait for me to say 'next' before you continue, and stop for my approval before anything irreversible such as sending, submitting, deleting, or paying.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Run a two-pass bug check before opening a PR
Use an agent to flag possible bugs first, then have it verify each one before any pull request is opened.
Auto-file Svelte mistakes as GitHub issues
They added an opt-in skill that turns concrete Svelte mistakes into draft or filed issues in the Svelte AI tools repo.
Create a CLI coding agent from a Python LLM base
Claude Code turned Simon Willison’s Python LLM template into a working CLI coding agent from two prompts.
Check traces for hidden agent errors
Laminar’s flow-1 reviews agent traces end to end so quiet mistakes can be caught without sampling.
Set up a GitHub bot that triages and patches bugs
They built a GitHub bot that sorts issues, reviews pull requests, answers with citations, and opens patch PRs for humans to check.
Find native browser replacements in a web app
A coding-agent skill helps developers swap older front-end patterns for newer browser APIs.