diff --git a/dashboards/tracker-web/docs/TEST_VALIDATION_LOG.md b/dashboards/tracker-web/docs/TEST_VALIDATION_LOG.md
new file mode 100644
index 00000000..8c48ff23
--- /dev/null
+++ b/dashboards/tracker-web/docs/TEST_VALIDATION_LOG.md
@@ -0,0 +1,60 @@
+# Test Validation Log — `@bytelyst/tracker-web`
+
+This log tracks the validation/hardening effort for the tracker-web dashboard's test
+suite. Every gate is run from the repo root, filtered to this package. A gate is only
+marked green if it was actually executed and passed.
+
+All commands are run as:
+
+```bash
+cd /Users/sd9235/code/mygh/learning_ai_common_plat
+pnpm --filter @bytelyst/tracker-web <script>
+```
+
+---
+
+## Baseline (before any changes)
+
+Recorded at repo HEAD `77b074f3`.
+
+| Gate          | Command                                             | Result | Notes                                                                                  |
+| ------------- | --------------------------------------------------- | ------ | -------------------------------------------------------------------------------------- |
+| typecheck     | `pnpm --filter @bytelyst/tracker-web typecheck`     | PASS   | `tsc --noEmit`, exit 0                                                                 |
+| lint          | `pnpm --filter @bytelyst/tracker-web lint`          | PASS   | `eslint`, exit 0                                                                       |
+| test (unit)   | `pnpm --filter @bytelyst/tracker-web test`          | PASS   | 8 files, 55 tests                                                                      |
+| test:coverage | `pnpm --filter @bytelyst/tracker-web test:coverage` | RUNS   | 90.78% stmts / 84.78% branch (only files imported by tests are measured)               |
+| build         | `pnpm --filter @bytelyst/tracker-web build`         | PASS   | `next build --webpack`, 14 routes                                                      |
+| format:check  | `pnpm --filter @bytelyst/tracker-web format:check`  | FAIL   | `README.md`, `docs/roadmaps/UX_INTEGRATION_BYTELYST.md` not Prettier-formatted         |
+| test:e2e      | `pnpm --filter @bytelyst/tracker-web test:e2e`      | FAIL   | 9 passed / 3 failed (see below)                                                        |
+| size:check    | `pnpm --filter @bytelyst/tracker-web size:check`    | FAIL   | `bundlesize` cannot load — `iltorb` native module won't build on Node 25 (env blocker) |
+
+### Baseline e2e failures (3)
+
+1. `Tracker Login Page › shows error for invalid credentials` — depends on a live
+   platform-service. With no backend, the login proxy returns 502
+   "Platform service unavailable", which does not match `/failed|error|invalid/i`.
+2. `Tracker — Public Roadmap › roadmap page can toggle between board and list view` —
+   list view renders "No items found" (no backend data), so `table, [role='list']`
+   never appears.
+3. `Tracker — Health › GET /api/health returns ok` — the dev server has no
+   `PLATFORM_API_URL` / `JWT_SECRET` / `DEFAULT_PRODUCT_ID` set, so `/api/health`
+   correctly returns `degraded` (503), not `ok`.
+
+Root cause for all three: the e2e suite depends on a live platform-service and ambient
+env, which violates the determinism requirement ("mock network at the route/proxy
+boundary; don't depend on a live platform-service").
+
+### Out-of-scope / environment blockers
+
+- **`size:check` (bundlesize)**: `bundlesize@0.18.2` (latest, unmaintained) transitively
+  requires `brotli-size@0.1.0` → `iltorb@2.4.5`, a deprecated native addon that does not
+  compile on Node 25 (`Cannot find module './build/bindings/iltorb.node'`). `pnpm rebuild
+iltorb` produces no `build/` output. The broken dependency lives in the root
+  `node_modules` (shared infra, out of scope to patch), and no newer `bundlesize` exists.
+  Bundle budgets are instead verified manually against the gzipped build output.
+
+---
+
+## Changes
+
+(Subsequent entries appended below as work proceeds.)