Skip to content
agentusehub

Audit hundreds of pull requests with a terminal coding agent

In an independent test, Muse Code audited 222 pull requests in under five minutes for about ten cents, but also built a plan on a Google project that doesn't exist.

Time saved
Not reported
Setup time
Not reported
Running cost
Not reported
Cost range
Not reported

Use case overview

In an independent test, Muse Code audited 222 pull requests in under five minutes for about ten cents, but also built a plan on a Google project that doesn't exist.

Reported experience

  1. What they did

    The article only summarizes this test and doesn't give its setup. It describes the most credible independent hands-on run as Muse Code, Meta's terminal agent, auditing 222 pull requests. Separately, the agent researched a Google project for three minutes. The project doesn't exist, and the agent built its whole plan on it. The guide covers installation, the default sandbox and approval prompts, and parallel subagents in git worktrees.

  2. How it went

    The audit of 222 pull requests finished in under five minutes for about ten cents. The long-horizon planning run failed outright on the fabricated project, which the article says is where the agent is weakest.

  3. Worth knowing

    It's a beta. The cheap $0.10/$0.20 tier trains on your code and is limited to 100 requests per minute, so use the standard tier for anything proprietary.

Example prompts

Try with Meta Muse
Using the GitHub CLI, audit the open pull requests in [owner/repo]. For each one, summarize what it changes, flag stale, duplicate, or risky PRs, and give a suggested action. Only reference things you can verify exist in the repo, and say so when you are unsure. Finish with a ranked summary table and do not comment on, merge, or close anything without showing me first.
Software & DevOps

Port a Chrome extension to Safari

The author had ChatGPT take the Session Buddy extension source from Chrome, port it to Safari and install it, taking about 20 minutes.

ChatGPT