docs: add structured chat readiness report

structured-chat-readiness-report.md

@@ -0,0 +1,42 @@
+# Structured native-chat readiness report
+
+Integration branch: `brennanb2025/structured-chat-integration-latest`  
+Integration head: `3db8e822e45fa3623feb610a89e37eb2744fa596`  
+Base recorded: `origin/main` = `02a7742406a5a84fb372d6255d5a4367421990bd`
+
+## Exact implementation heads
+
+| Unit | Branch / head | Recommendation |
+| --- | --- | --- |
+| Claude/provider | `brennanb2025/claude-structured-current` / `ce8e5b5ab6ca0102375ff0765d4a1e3ce8d6ab88` | land after UI proof and independent PR review |
+| Windows | `OrcaWin/windows-structured-current` / `f8e351ebaeb2bfd7ec93893ea75f30226572fa97` | land after path-length CI and real-host validation |
+| WSL | `brennanb2025/wsl-structured-local` / `a690b67aaa` | land after selected-distro validation |
+| SSH/Linux | `brennanb2025/ssh-structured-current` / `e4aa13d4cc` | land after real SSH/OpenClaw validation |
+| Paired remote | `brennanb2025/remote-structured-current` / `736e01d8d7` | land after paired-host and cross-version evidence |
+| Linux packaging | `brennanb2025/linux-packaging-structured-split` / `b96801e9aa` | do not land as-is; ancestry is broad and needs a focused re-cut |
+| Combined integration | `brennanb2025/structured-chat-integration-latest` / `3db8e822e4` | land after the platform evidence gates below |
+
+## Verification completed
+
+- `pnpm test` main/Claude/Codex/WSL/SSH suites: **49 files, 414 passed, 2 skipped**.
+- Renderer routing suites: **3 files, 24 passed**.
+- `pnpm tc:node`: passed.
+- `pnpm tc:web`: passed.
+- Changed-file `oxlint` (including React doctor config): passed.
+- `ref-oss` synchronization completed before implementation/review.
+- Architecture split recorded in `docs/architecture/structured-native-chat-platform-split.md`.
+
+## Platform evidence
+
+- **Windows:** worker ran on Windows high-spec Orca host; focused tests and lint passed. Full validation is blocked by node-gyp/MSB3491 path-length failure (>260 characters), so no packaged Windows proof is claimed.
+- **WSL:** selected-distro argv/path/WSLENV and ownership tests pass. The Windows remote runtime could not complete a worker attach (timeout, then disk-full on alternate runtime); actual distro/provider and Electron proof remain unverified.
+- **Native Linux:** packaging contracts and bounded smoke coverage exist in the prior branch, but the worker branch mixed provider/WSL/Codex changes and is excluded from integration. Ubuntu 20.04/arm64 packaged execution was not run here.
+- **SSH/Linux:** verdict tests pass and preserve exactly `live`, `unverifiable`, `exited`. OpenClaw was unreachable during validation; no real disconnect/reconnect run is claimed.
+- **Paired remote Orca:** host-aware routing, capability projection, and unsubscribe cleanup tests pass. No paired-host Electron run was available; cross-version harness is covered by existing unit tests but requires release-runner confirmation.
+- **Electron UI:** the Claude worker attempted macOS Electron/CDP, but the endpoint became unresponsive. Visible UI proof is therefore unverified; no computer-use or OS-level automation was used.
+
+## Readiness/review outcome
+
+The implementation and platform workers each ran their requested readiness/review scans. The coordinator reran focused tests, typechecks, and changed-code lint on the combined head. No proven in-scope correctness issue remains in those checks. Remaining evidence blockers are infrastructure/topology gates, not silently passed support claims; until they are run, recommendations remain “land after fixes” for affected units.
+
+Legacy behavior is preserved when the structured toggle is off, SSH transport loss remains inconclusive, unknown prompt outcomes are not resent, and remote execution never starts a local provider.