docs: final wrap-up sync — accurate test counts + roadmap links

CLAUDE.md

@@ -88,10 +88,12 @@ pnpm --filter @gitmarks/extension-chrome e2e
 1. ✅ `@gitmarks/core`
 2. ✅ Chrome MVP (toolbar save)
 3. ✅ Chrome native tree integration
-4. ⬜ Firefox build (`webextension-polyfill` over the same source)
-5. ⬜ Web UI: list / search / tag management (Vite + React + Tailwind SPA)
-6. ⬜ Web UI: bulk operations + trash + export
-7. ⬜ Safari (`safari-web-extension-converter` over the Chrome bundle)
+4. ⬜ Firefox build (`webextension-polyfill`) — issue [#23](https://github.com/paperhurts/gitmarks/issues/23)
+5. ⬜ Web UI v1: list / search / tag management — issue [#24](https://github.com/paperhurts/gitmarks/issues/24)
+6. ⬜ Web UI v2: bulk operations + trash + export — issue [#25](https://github.com/paperhurts/gitmarks/issues/25)
+7. ⬜ Safari (`safari-web-extension-converter`) — issue [#26](https://github.com/paperhurts/gitmarks/issues/26)
+
+For next-piece-of-work: pick one of #23–#26. Each has a scope block in its issue description. The plan-driven workflow (`docs/superpowers/plans/YYYY-MM-DD-<feature>.md`) is the expected approach for anything larger than ~3 commits.
 
 ## Non-goals (do not implement)
 

README.md

@@ -130,11 +130,14 @@ The load-bearing invariants:
 ## Files in this repo
 
 - `spec.md` — full design spec (source of truth for non-obvious decisions)
+- `CONTRIBUTING.md` — branch/PR conventions, TDD policy, plan-driven workflow
 - `CLAUDE.md` — guidance for AI agents working in this repo
+- `LICENSE` — MIT
 - `docs/superpowers/plans/` — implementation plans, one per branch
 - `packages/*/README.md` — package-specific documentation
 - `examples/example-bookmarks-repo/` — sample `bookmarks.json` + `tags.json`
   to seed a fresh repo, used by `@gitmarks/core` fixture tests
+- `.github/workflows/test.yml` — CI (typecheck + unit tests + build on every PR)
 
 ## Contributing
 

packages/extension-chrome/README.md

@@ -168,19 +168,23 @@ pnpm --filter @gitmarks/extension-chrome typecheck
 
 **Coverage:**
 
-Unit tests (53) cover the pure logic — settings, machine ID, bookmark
+Unit tests (97) cover the pure logic — settings, machine ID, bookmark
 factory, save flow, folder path conversion, ID mapping, suppression
-registry, apply-remote, reconciliation, and the listener
-batch/debounce/flush algorithm. Tests use a vitest setup file that stubs
+registry (URL + node ID), apply-remote, reconciliation, the listener
+batch/debounce/flush algorithm, and the background-core poll/reconcile
+orchestration. Tests use a vitest setup file that stubs
 `chrome.storage.local` and `chrome.bookmarks.*` with an in-memory backend.
 
-E2e tests (6) launch real Chromium with the built extension, mock the
-GitHub API via Playwright route interception, and exercise:
-- Popup before-setup → "Set up gitmarks" button visible
-- Options page validate + save flow with the mocked API
-- "Save this page" round-trip → mocked GitHub receives the PUT
-- `chrome.bookmarks.create` → expected PUT payload
-- Remote add seeded into the mock → bookmark appears in local tree
+E2e tests (4 passing, 2 skipped) launch real Chromium with the built
+extension and mock the GitHub API via Playwright route interception:
+- ✓ Popup before-setup → "Set up gitmarks" button visible
+- ✓ Options page validate against missing `bookmarks.json` (404 path)
+- ✓ `chrome.bookmarks.create` → expected PUT payload
+- ✓ Remote add seeded into the mock → bookmark appears in local tree
+- ⊘ Save flow via popup → GitHub PUT (skipped: depended on the dropped
+  `tabs` permission fallback; the save-flow logic is fully covered by
+  unit tests)
+- ⊘ Popup save view after configure (same dependency, same coverage)
 
 **Known gap (not fully understood):** during development we couldn't
 reliably get `chrome.bookmarks.*` events dispatched from