Improve content at 4 anchor points for better app-link UX

- set-up-a-machine/overview.md #sbc-setup-instructions: replace generic
  tip box with direct links to supported SBC setup guides
- data/capture-sync/capture-and-sync-data.md #supported-resources: add
  table of component types and capturable methods (verified against
  rdk/components/*/collectors.go)
- train/train-a-model.md #upload-your-training-script: add upload steps
  with CLI command instead of bare pointer link
- train/train-a-model.md #click-for-more-information-on-parsing-command-
  line-inputs: add the actual flags (--dataset_file, --model_output_directory,
  custom --key=value args) verified against rdk/cli/ml_training.go

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: shannonbradshaw

docs/data/capture-sync/capture-and-sync-data.md

@@ -90,7 +90,24 @@ If you see data flowing in, capture and sync are working correctly.
 
 ## Supported resources {#supported-resources}
 
-Any component that implements a Viam API method returning data can be configured for data capture. This includes cameras, sensors, movement sensors, encoders, and any modular component that returns readings. See the [data reference](/data/reference/) for the full list of supported methods and data types.
+Any built-in or modular component that implements a capturable method can be configured for data capture. The following table lists the most commonly captured component types and their methods:
+
+| Component type | Capturable methods |
+|---|---|
+| Camera | `ReadImage`, `GetImages`, `NextPointCloud` |
+| Sensor | `GetReadings` |
+| Movement sensor | `GetPosition`, `GetLinearVelocity`, `GetAngularVelocity`, `GetCompassHeading`, `GetLinearAcceleration`, `GetOrientation`, `GetReadings` |
+| Encoder | `GetTicksCount` |
+| Motor | `GetPosition`, `IsPowered` |
+| Power sensor | `GetVoltage`, `GetCurrent`, `GetPower`, `GetReadings` |
+| Servo | `GetPosition` |
+| Arm | `GetEndPosition`, `GetJointPositions` |
+| Gantry | `GetPosition`, `GetLengths` |
+| Board | `Analogs`, `Gpios` |
+| Vision service | `CaptureAllFromCamera` |
+| SLAM service | `GetPosition`, `GetPointCloudMap` |
+
+Every component and service also supports capturing `DoCommand` responses. Modular components that return readings through any of these methods are automatically capturable.
 
 ## Troubleshooting
 

docs/set-up-a-machine/overview.md

@@ -56,12 +56,16 @@ Use the **Platform you want to run on** dropdown to select the operating system
 
 Options include Linux / Aarch64, Linux / x86, Mac, Windows native, Windows (WSL), Linux / Armv7l, and ESP32.
 
-{{< alert title="Tip" color="tip" >}}
+If you're using a single-board computer, follow the setup guide for your board before continuing:
 
-If you're using a single-board computer like a Raspberry Pi or NVIDIA Jetson, make sure it's running a compatible Linux OS before continuing.
-The setup page links to an [installation guide](/reference/device-setup/) for supported single-board computers.
+- [Raspberry Pi](/reference/device-setup/rpi-setup/)
+- [NVIDIA Jetson Nano](/reference/device-setup/jetson-nano-setup/)
+- [NVIDIA Jetson AGX Orin](/reference/device-setup/jetson-agx-orin-setup/)
+- [BeagleBone AI-64](/reference/device-setup/beaglebone-setup/)
+- [Orange Pi Zero 2](/reference/device-setup/orange-pi-zero2/)
+- [Orange Pi 3 LTS](/reference/device-setup/orange-pi-3-lts/)
 
-{{< /alert >}}
+See [all supported boards](/reference/device-setup/) for the full list.
 
 ## 4. Select your installation method
 

docs/train/train-a-model.md

@@ -266,11 +266,28 @@ for _, job := range jobs {
 
 ## Upload your training script {#upload-your-training-script}
 
-To write and upload a custom training script instead of using a built-in training type, see [Custom training scripts](/train/custom-training-scripts/#package-and-upload).
+To use your own training script instead of a built-in training type:
+
+1. Package your script with a `setup.py` or `pyproject.toml`. Viam invokes it as `python3 -m model.training`.
+2. Upload with the CLI:
+
+   ```sh {class="command-line" data-prompt="$"}
+   viam training-script upload --path=<path-to-script> --org-id=<org-id> --script-name=<name>
+   ```
+
+3. Submit a training job that references your script (from the web UI or CLI).
+
+For the full walkthrough including local testing with Docker, see [Custom training scripts](/train/custom-training-scripts/).
 
 ## Parse command line inputs {#click-for-more-information-on-parsing-command-line-inputs}
 
-For details on parsing command line arguments in custom training scripts, see [Write a training script](/train/custom-training-scripts/#trainingpy).
+When you submit a training job with custom arguments, Viam passes them to your training script as command-line flags. Your script receives:
+
+- `--dataset_file`: path to the dataset manifest (default: `dataset.jsonl`)
+- `--model_output_directory`: where to write the trained model (fixed: `/model_output`)
+- Any custom arguments you defined, as `--key=value` pairs
+
+Argument keys must be alphanumeric with underscores or hyphens. For full details on writing and testing training scripts, see [Custom training scripts](/train/custom-training-scripts/).
 
 ## What's next