CI: gate on Prettier; docs: disclose AI assistance

- ci.yml runs `npm run format:check` before the build.
- CONTRIBUTING / AGENTS: point at `npm run format` instead of the old
  "no formatter" note.
- README: new "Built with AI assistance" section — the project is
  developed with Claude / Claude Code, human-directed and reviewed;
  visible in the Co-Authored-By trailers. CONTRIBUTING asks contributors
  to disclose AI-assisted PRs.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017Ve7HLspzeG2xDPtJQ8vmu

.github/workflows/ci.yml

@@ -21,6 +21,8 @@ jobs:
 
       - run: npm ci
 
+      - run: npm run format:check
+
       # The build's `prebuild` step fetches a data snapshot from PokéAPI.
       # Cache it so only the first run (or a change to the script) pays that
       # cost; CI only needs *a* snapshot to prove the build compiles.

AGENTS.md

@@ -21,9 +21,10 @@ npm run preview         # serve the production build — REQUIRED to test the se
 ```
 
 - Node **20+**.
-- There is **no lint or test command** and **no test suite**. The build
-  (`npm run build`) is the only automated check — it must pass. It runs
-  esbuild over every module, so it catches syntax and import errors.
+- Automated checks (both run in CI, both must pass):
+  `npm run format:check` (Prettier) and `npm run build` (esbuild transforms
+  every module, so it catches syntax and import errors). Run `npm run format`
+  before finishing. There is **no test suite** and no ESLint.
 - `npm run snapshot -- --force` rebuilds the snapshot even if it's fresh.
 
 ## Project structure
@@ -60,8 +61,8 @@ npm run preview         # serve the production build — REQUIRED to test the se
   caught set.
 - Respect `prefersReducedMotion()` (from `src/store/settings.js`) for any
   animation.
-- 2-space indent, semicolons, single quotes, trailing commas in multi-line
-  literals. Match the surrounding file; keep comments about _why_.
+- Formatting is Prettier (`npm run format`). Otherwise match the
+  surrounding file; keep comments about _why_.
 - No new runtime dependencies. Prefer adding to the snapshot over a new
   runtime fetch.
 
@@ -90,7 +91,8 @@ npm run preview         # serve the production build — REQUIRED to test the se
 
 ## Verifying a change
 
-`npm run build` must pass. Then sanity-check by hand: the dex grid
+`npm run format` then `npm run build` — both must pass. Then sanity-check
+by hand: the dex grid
 (filters, sort, catching from a card), switching games (numbers + typings
 update), a detail page (all tabs, form switch), and `npm run preview`
 offline (DevTools → Network → Offline → reload). Check light and dark

CONTRIBUTING.md

@@ -54,11 +54,13 @@ See the **Project layout** and **How it works** sections in
 
 ## Code style
 
-There's no linter or formatter config yet — **match the surrounding
-code**:
+Formatting is **Prettier** (`.prettierrc.json`). Run `npm run format`
+before committing; CI runs `npm run format:check` and will fail on
+unformatted code. There's no ESLint — for everything Prettier doesn't
+decide, **match the surrounding code**:
 
-- 2-space indent, semicolons, single quotes, trailing commas in
-  multi-line literals (an `.editorconfig` covers whitespace).
+- 2-space indent, semicolons, single quotes, 100-col width, trailing
+  commas (all handled by Prettier).
 - Build DOM with the `el()` helper from `src/lib/dom.js`, not template
   strings or `innerHTML` (except the deliberate `html:` prop for trusted
   inline SVG).
@@ -110,6 +112,10 @@ Describe what you tested in the PR.
 - Fill in the PR template. Screenshots or a short clip for anything
   visual.
 - It's fine to open a draft PR early to discuss direction.
+- AI-assisted contributions are welcome — say so in the PR description if a
+  tool wrote a meaningful part of it, and review it yourself first. The
+  same bar applies either way. This project is itself built this way (see
+  the README).
 
 ## Reporting bugs
 

README.md

@@ -275,6 +275,21 @@ coordinates), no automated test suite yet, and no formal Lighthouse pass.
 
 ---
 
+## Built with AI assistance
+
+Pocketdex is developed with heavy use of AI coding tools — primarily
+Anthropic's Claude, via Claude Code. Direction, design decisions, review
+and testing are human-led; a large share of the implementation is
+AI-generated and then reviewed before it lands. You'll see this in the git
+history (`Co-Authored-By: Claude …` trailers).
+
+This doesn't change anything for contributors: use whatever tools you like,
+disclose AI-generated PRs in the description, and expect the same review
+either way. There's an [AGENTS.md](AGENTS.md) so agents have the house
+rules too.
+
+---
+
 ## Legal
 
 Pocketdex is an unofficial, non-commercial fan project. It is **not