Who it is forA developer who just had an AI agent write a feature and wants a second look before committing.
The problemAgent-written code often adds more code than the task needs, and bugs or missing tests are easy to miss in the diff.
The outcomeA list of findings, each saying what the code does, what goes wrong, how to fix it, and what happens if it is not fixed.
What you need first
- Claude Code or Codex installed
- A git repository with a project to work on
- Permission to install a plugin into your agent
Steps
- Run
/plugin marketplace add DietrichGebert/ponytailin Claude Code. - Run
/plugin install ponytail@ponytailin Claude Code. - Let the agent make your change in the repo.
- Run
/ponytail-review uncommittedto review the current uncommitted changes. (inferred, check the README) - Read each finding and fix the ones that matter before committing.
Prompt for your agent
In the repo <repo-path> (DietrichGebert/ponytail skill installed), run /ponytail-review uncommitted on the current changes. Expected output: a list of findings with what the code does, what goes wrong, how to fix it, and the cost of not fixing it. Verify that each finding points to a file touched by the uncommitted diff.
Caveat. The README says the review is meant to catch bugs and missing tests, but it does not promise it will find every problem, so keep your own tests and checks.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for DietrichGebert/ponytail and "Review agent-written changes before merging"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer