diff --git a/docs/testing/pr-testing-guide.md b/archive/2026-06-test-cleanup/docs/testing/pr-testing-guide.md
similarity index 100%
rename from docs/testing/pr-testing-guide.md
rename to archive/2026-06-test-cleanup/docs/testing/pr-testing-guide.md
diff --git a/docs/testing/test-organization.md b/archive/2026-06-test-cleanup/docs/testing/test-organization.md
similarity index 100%
rename from docs/testing/test-organization.md
rename to archive/2026-06-test-cleanup/docs/testing/test-organization.md
diff --git a/docs/testing/testing-architecture.md b/archive/2026-06-test-cleanup/docs/testing/testing-architecture.md
similarity index 100%
rename from docs/testing/testing-architecture.md
rename to archive/2026-06-test-cleanup/docs/testing/testing-architecture.md
diff --git a/docs/testing/README.md b/docs/testing/README.md
new file mode 100644
index 00000000..3f05e414
--- /dev/null
+++ b/docs/testing/README.md
@@ -0,0 +1,73 @@
+# Testing
+
+GraphDone has one reproducible test entry — the **unified harness** — plus the
+focused **smoke gate** that must pass before any "it works" claim.
+
+```bash
+TEST_URL=http://localhost:3127 npm run test:smoke   # THE GATE (blocking, ~30s)
+npm run test:unified                                # full battery → dual report
+npm run test:unified:smoke                          # fast subset (unit + gate + a focused e2e)
+npm run test:unified:open                            # full battery, then open the HTML report
+npm run test:unified:lib                             # the harness's own unit tests (node --test)
+```
+
+## Dual output (one run, two artifacts)
+The harness writes to `test-artifacts/unified/`:
+- **`report.html`** — human-readable: per-sequence pass/warn/fail with embedded
+  `<img>` screenshots and `<video>` `.webm` clips (from the showcase project),
+  plus the cloud-audit summary. To share it, zip the whole folder (assets are
+  referenced relatively).
+- **`report.json`** — machine-parsable (`schema: graphdone.unified-report/1`):
+  rollup status + totals + per-sequence counts. Per-sequence raw JSON lands in
+  `test-artifacts/unified/sequences/`.
+- **exit code** = rollup status (any blocking failure → 1), so it doubles as a CI gate.
+
+## How it's wired
+- `tests/run-unified.mjs` — the entry. `--profile {smoke|pr|full|report}`,
+  `--sequence <id>`, `--open`, `--out <dir>`, `--cloud-findings <path>`.
+- `tests/sequences/unified.config.mjs` — declarative `SEQUENCES` + `PROFILES`.
+  Each sequence names an adapter (`vitest` | `playwright` | `cloud-audit`) + inputs.
+- `tests/lib/` — the reusable layer:
+  - `reporting/` — `aggregate.mjs` (pure rollup, unit-tested), `html.mjs`, `json.mjs`,
+    and the `generate-*-report.mjs` domain generators.
+  - `adapters/` — `playwright.mjs` (parses PW JSON + harvests video/screenshots),
+    `vitest.mjs`, `cloud-audit.mjs` (ingests the sibling GraphDone-Cloud live-audit).
+  - `runner/runSequence.mjs`, plus the shared helpers (`auth.ts`, `api.ts`, …) and
+    `metrics/balance_metrics.py`.
+
+## Profiles
+| profile | sequences | use |
+| --- | --- | --- |
+| `smoke` | unit-web, smoke gate, a focused auth e2e | fast blocking signal |
+| `pr` | + graph e2e, mobile, perf budgets | broader blocking set for PRs |
+| `full` | + diagnostics, showcase (video), matrix, vlm, perf-scale, cloud-audit | everything |
+| `report` | the capture-heavy report sequences only | regenerate galleries |
+
+## Test tree
+- `tests/e2e/<domain>/` — `smoke` (THE GATE), `auth`, `graph`, `ui`, `a11y`,
+  `mobile`, `responsive`, `api`, `admin`, and `reports/` (capture-heavy, own
+  Playwright projects, excluded from the fast default project).
+- `tests/diagnostics/<concern>/` — `layout`, `interactions`, `hierarchy`,
+  `inspector`, `physics`, `perf`, `ui` (the `geometry` project; report-only).
+- `tests/perf/` — ADAPT-8 budget gate + the large-scale sweep.
+- `tests/integration/` — infra specs (TLS, installation), out of the default project.
+- One canonical `playwright.config.ts` at the repo root.
+
+## Adding tests
+- **A new e2e spec:** drop it in the right `tests/e2e/<domain>/` folder; import
+  helpers via `../../lib/<helper>` (depth-2). It's auto-collected by the default
+  project (tag `@smoke`/`@core`/`@mobile` as appropriate).
+- **A new harness sequence:** add an entry to `tests/sequences/unified.config.mjs`
+  and list it in the relevant profile(s). Blocking sequences belong in `smoke`/`pr`.
+- **The smoke gate** lives at `tests/e2e/smoke/user-smoke.spec.ts` — keep it the
+  honest "a real user can use the app" check.
+
+## Authentication in tests
+Use the battle-tested helper for every authenticated spec:
+```ts
+import { login, navigateToWorkspace, TEST_USERS } from '../../lib/auth';
+```
+
+> Retired/legacy test files (and the previous `run-all-tests.js` / `run-pr-tests.js`
+> runners these guides used to describe) are preserved under
+> `archive/2026-06-test-cleanup/` with a MANIFEST — mine them before deleting.