### Source Peer feedback, collected in #167. Quotes (translated): use AI to check the pull requests, so it could have caught leftovers of the Vite template and titles that become invisible in the dark theme; let AI compare the documentation with the actual code and build output at every PR; a last round that compares README and docs with the app on the VM, runs Prettier on the whole code base and analyses the bundle; use AI to suggest tests of whole user flows and to generate tests of edge cases and review accessibility. ### What to do - [ ] Write a short checklist for a pull request that the author runs with AI before asking for review: leftovers (unused files and imports), the dark theme of every new component, docs that mention what the PR changes, `npm run build` size against the README, Prettier. - [ ] Put the checklist in the pull request template or in the README "Working in this repository". - [ ] Do one round now: compare README and `docs/` with the code and with the app on the VM, and list the differences (the fixing is in #148). - [ ] Ask the AI for edge cases and whole user flows that the tests do not cover, and add the useful ones (see #158). ### Done when - [ ] The checklist is in the repository, one round is done, and the findings are linked. _Comment made with Claude Code in discussion with Ronja_