# What it finds

> Every kind of finding Excellent shows, what each button on a finding does, which fixes can't be undone, and what Explain sends.

Each finding says what happened, which agent did it, where and when, with up to three examples. Your email and keys
are masked in what it shows you.

## The findings

| Finding | What it means |
| --- | --- |
| Your email was sent to websites | An agent put your address in a web request: a header, a web address or a form post. |
| Your email was typed into web forms | A browser agent typed your address into a form on a website. |
| API keys are sitting in your agent logs | A key that looks real was typed into a command, so it's stored in plain text in a log. |
| Agents opened keys and secrets files | A command read SSH keys, `.env` files or your keychain. |
| History was overwritten | An agent force pushed. |
| Safety checks were skipped | A commit or push used `--no-verify`, which skips your git hooks. |
| Never-delete folders were touched | An agent deleted or moved a folder you marked never-delete. |
| Your own rules were broken | An agent did something a rule in your `CLAUDE.md` or `AGENTS.md` says not to do. |
| Things were published or deployed | An agent published a package or deployed to production. Check they were all meant. |
| Some of your safety checks aren't running | A hook in an agent's settings points at a program that's gone. |
| Commits were made without your repo's checks | In a repository whose commits Excellent watches, a commit had no passing check. See [Git protection](https://www.excellent.so/docs/git). |

Separately, the menu shows the rules you gave only one of Claude Code and Codex, under **Not told to every agent**.

## What the buttons do

Most findings are about what already happened. With Pro, their main button stops it next time:

- **Block this from now on** or **Ask me first from now on** turns on that rule of Excellent's check. If protection
  isn't on yet, it turns it on for your agents. These need Pro. See [Turn on protection](https://www.excellent.so/docs/turn-on-protection).
- **I meant this** or **That's fine** hides the finding.

Three kinds of finding have a real fix:

| Button | What it does | Undo |
| --- | --- | --- |
| **Remove keys from the logs** | Replaces each key in your agents' logs with a marker, in place. | Can't be undone. Excellent asks first. The key still works until you replace it with its provider. |
| **Remove it from N files** | Asks your own Claude Code or Codex to take your email out of scripts it wrote, up to three files at a time. Excellent then reads each file again to check. | Your agent edits the files; there's no backup. Excellent asks first, every time. |
| **Repair N hooks** | Points hooks whose Node version is gone at the Node you have now. Offered when the missing Node came from nvm and another nvm Node is installed. | Each settings file is backed up first. |

Rules that reach only one agent have a **Share…** button in the menu. See [Your rules, shared](https://www.excellent.so/docs/rules).

> **Remove it from N files** gives your agent the file's path. Your agent reads the file itself, so what's in it,
> including your email, goes to its provider under your account, like any work you give it.

## Explain

**Explain** asks your own Claude Code (or Codex, if you don't have Claude Code) to explain a finding in plain words.
It hands over the finding's title, summary, count, dates and up to three examples, with your email and keys in a
known format masked first. Your agent sends that to its provider, on your plan. It runs without tools that edit
files. Excellent asks first; you can tell it not to ask again.

Source: https://www.excellent.so/docs/what-it-finds
