# A preference of mine I didn't recognise

My site's repo had a note for coding agents that said I didn't want them taking screenshots:

> Ibrahem's box is WSL2 with no usable display path, and he does not want that route attempted.

It had a reason, "He asked for code review, not visual QA", and instructions: review visual changes by reading the CSS, and reason about rendering "from the CSS and the SVG path data directly."

When a session found the note at the end of October, I didn't recognise it.

## Where it came from

It went into the repo at the start of October, with five other notes, when I set the project up for cloud sessions. A later session's log called all six auto-memory leftovers, and this note's frontmatter says `type: feedback`.

So most likely a session wrote down that I didn't want screenshots and filed it as feedback from me. Nothing records who wrote it. Another note in the same set already treated it as a rule: `[[no-browser-screenshots-wsl]] applies`.

## There was a display

The note said there was no working display. There was. Headless `google-chrome --screenshot` and the chrome-devtools MCP both rendered the live site on this machine, desktop in light mode and mobile in dark, right to left. The note went into the archive.

It wasn't the only stale line. The same session found `CLAUDE.md` describing a nav the site didn't have.

## Where rules live

Auto-memory is off for me, everywhere. A rule that changes how a session works goes in the project's `CLAUDE.md`.

My global config tells every session how to treat the rest. Docs, notes, comments, a project's state file and a session's own earlier answers are maps: they say where to look, never what's true.
