I’m building a tiny CLI that tells you what your coding agent changed and which edits are risky. What would you want it to flag?
retained text
I use coding agents (Claude Code, Codex, Cursor) every day at work. The part that’s started to bother me isn’t the code they write, it’s reviewing it afterward. After a long run, I get a 600-line diff across 14 files. Most of it is fine. But somewhere in there it quietly edited a config, deleted a test, or touched auth code I didn’t ask it to. I either read every line or I just trust it, and both feel wrong. So I’m building a small open-source CLI called **lineup**. v1 does two things: * `lineup diff`: a plain-English summary of what the agent changed, grouped by intent instead of by file * `lineup risky`: flags edits worth a second look (deleted tests, config/env changes, auth or payment code, new dependencies, files outside the task) It’s not a PR reviewer. CodeRabbit reviews your PR. lineup tells you what your agent just did, before it ever becomes a PR. Building it this week, then a few teammates will use it for real. Two questions before I go too far: 1. What’s the edit an agent made that you’d most want flagged? 2. How do you review agent changes today: read every line, git diff, or just trust it?
document fetched 2026-11-02 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1xl9ysa/im_building_a_tiny_cli_that_tells_you_what_your