Add bidirectional RF handoff coordination

docs/handoff/RX_AGENT_USAGE.md

@@ -1,76 +1,83 @@
 # RX watcher usage
 
-## START
+## Start
 
-On the RX laptop, pull `video-development`, install the project's normal Python dependencies, then double-click:
+On the RX laptop, pull video-development and double-click:
 
-```text
-tools\start_rx_watcher.cmd
-```
+    tools\start_rx_watcher.cmd
 
-The watcher stays in a visible console and polls `origin/video-development` every 30 seconds. The interval is configurable with `--poll-interval`. It does not start ChatGPT, Codex or another agent.
+The watcher stays in a visible console and polls origin/video-development every 30 seconds. It does not start ChatGPT, Codex or another agent.
 
-## IDLE
+## Idle
 
-While no new `TASK_ID` exists, the watcher performs only `git fetch`, reads the remote task file and compares IDs with `.local/rx_agent_state.json`. There is no model execution or token use.
+While no new TASK_ID exists, the watcher performs only git fetch, reads the remote task file and compares IDs with .local/rx_agent_state.json. There is no model execution or token use.
 
-## NEW TASK
+## New task
 
 The console shows:
 
-```text
-NEW TASK DETECTED
+    NEW TASK DETECTED
 
-TASK ID =
-TITLE =
-SAFETY LEVEL =
-HARDWARE ACCESS =
-RF RX =
-RF TX =
-```
+    TASK ID =
+    TITLE =
+    SAFETY LEVEL =
+    HARDWARE ACCESS =
+    RF RX =
+    RF TX =
 
-Choose `[A] Approve`, `[D] Defer` or `[S] Stop watcher`. An RF TX task also requires the exact separate phrase `AUTHORIZE RF TX`; another RED task requires `AUTHORIZE <TASK_ID>`. This only opens the executor gate; the first version's `ManualExecutor` still performs no RF action and starts no model.
+Choose [A] Approve, [D] Defer or [S] Stop watcher. RF TX also requires the exact phrase AUTHORIZE RF TX. Another RED action requires AUTHORIZE <TASK_ID>.
 
-## APPROVE
+## Approve and execute
 
 After approval, open:
 
-```text
-.local\rx_agent_outbox\<TASK_ID>\TASK.md
-```
+    .local\rx_agent_outbox\<TASK_ID>\TASK.md
 
-Transfer that prepared task to the chosen agent manually. Fill the adjacent `RX_RESULT.md` and `RX_RESULT.json` with measured results. When work is actually complete, run:
+Transfer that prepared task to the chosen executor. The watcher records last_approved_task_id, which is required for result publication. Preparing the outbox is not evidence that the hardware task ran.
 
-```powershell
-python -B tools/rx_agent_watcher.py --mark-completed <TASK_ID>
-```
+When the task is actually complete, run:
 
-Do not claim completion merely because the prompt was prepared.
+    python -B tools/rx_agent_watcher.py --mark-completed <TASK_ID>
 
-## DONE and result publication
+## Publish READY
 
-If the task says `ALLOW_PUSH = RX_RESULT_ONLY` and the user separately authorizes publication, publish only small result Markdown/JSON and log metadata to `rx-node`. Never commit or push from the RX laptop to `video-development`. Do not put raw IQ, video or waveform files in Git.
+For RF001A_OTA_CAPTURE_002, create the strict READY JSON outside Git only after the capture child is alive and the file is growing. Then run:
 
-## STOP
+    python -B tools/rx_agent_watcher.py --publish-result <ready-json-path> --result-branch rx-node
+
+On success, the watcher prints RX RESULT PUBLISH = PUBLISHED or UNCHANGED, the result commit, and RF TX = 0.
+
+The command fails if:
+
+- the task was not approved locally;
+- the TASK_ID differs;
+- source_head differs from the checkout HEAD;
+- origin/rx-node does not exist;
+- the schema is invalid;
+- Git commit or push fails.
+
+A failed publication is not READY evidence and cannot authorize TX.
+
+## Publish COMPLETE
+
+After capture completion, build the strict final JSON with state CAPTURE_COMPLETE and publish it with the same command:
+
+    python -B tools/rx_agent_watcher.py --publish-result <final-json-path> --result-branch rx-node
+
+Only the two fixed result files are committed. Raw IQ, video, waveform, stdout, stderr and large logs stay outside Git.
+
+## Stop
 
 Double-click:
 
-```text
-tools\stop_rx_watcher.cmd
-```
+    tools\stop_rx_watcher.cmd
 
-The script creates a local stop request and waits for the watcher PID to clear. The approval prompt is interruptible on Windows, and the polling delay checks the stop request every 0.1 seconds. A successful stop prints `WATCHER STOPPED`; no polling process remains.
+The script creates a local stop request and waits for the watcher PID to clear. A successful stop prints WATCHER STOPPED.
 
-## DRY RUN
+## Dry run
 
 From the repository root:
 
-```powershell
-python -B tools/rx_agent_watcher.py --dry-run --once --approval approve --task-file tests/fixtures/watcher_test_task.md
-```
+    python -B tools/rx_agent_watcher.py --dry-run --once --approval approve --task-file tests/fixtures/watcher_test_task.md
 
-Expected behavior: the fake task is detected, approved and written under `.local/rx_agent_outbox/WATCHER_TEST_001/`. No Git fetch, SDR access, agent call, RF RX/TX, commit or push occurs. Delete `.local/` at any time if the local state and outbox are no longer needed.
-
-## Executor availability
-
-Version 1 always provides `ManualExecutor`. Before adding or selecting a Codex backend on RX, run `Get-Command codex` on that laptop and verify a harmless local invocation. Do not infer RX availability from the main PC.
+The fake task is written under .local/rx_agent_outbox/WATCHER_TEST_001/. No Git fetch, SDR access, agent call, RF RX, RF TX, commit or push occurs.