name	rokit
description	Use when testing, sideloading, deploying, or debugging Roku apps with rokit, including package/install/launch flows, remote key input, BrightScript console capture, ECP/SceneGraph/media-player observation, screenshots, readiness waits, and proof bundles.

rokit

Use this skill when a repo consumes @putdotio/rokit for Roku device harness work. Keep rokit generic and put app journeys, product selectors, fixture names, content IDs, account state, and product assertions in the consumer app repo.

Workflow

Run rokit describe to inspect the command surface instead of guessing flags or --input-json payload fields. Use rokit describe <command> when you only need one command schema.
Use --dry-run before mutating commands such as package, install, launch, press, screenshot, or proof.
Prefer structured output. In automation, rely on the non-TTY JSON default or pass --json explicitly.
Use --fields to keep observations small when only a few values are needed.
Screenshot commands append a timestamp to the requested filename. Use the returned JSON data.path as the artifact path instead of assuming the input path was written.
For live proof, use rokit snapshot for a quick state read, rokit proof <output-dir> for review artifacts, or pnpm live:probe in the rokit repo for the full generic package/install/launch/input/proof probe.
For crash/debug proof, start rokit console <output-path> before reproducing the problem so startup errors and crash traces are captured. Use rokit debug-command <command> only for generic Roku debug commands.
Use rokit wait-ready <app-id> [node-name] [condition] [value] after launch when the app can race ECP or SceneGraph readiness.
Use rokit press <keys...> --until-node <node> --until-state visible for bounded navigation loops instead of arbitrary sleeps.

Start Here

Read only the reference needed for the current task:

live proof, readiness waits, SceneGraph and media-player observation: references/live-proof.md
BrightScript console capture and debug-server commands: references/debugging.md

Common Commands

rokit describe
rokit describe proof
rokit --json snapshot --fields command,status,data.activeApp,data.mediaPlayer
rokit --dry-run package artifacts/live/channel.zip
rokit --dry-run launch dev
rokit --json proof artifacts/live/proof --screenshot
rokit wait-node title text "Ready"
rokit --input-json '{"command":"press","keys":["Down","Select"]}'
rokit --input-json @artifacts/rokit-payload.json
printf '{"command":"press","keys":["Back"]}' | rokit --input-json -
rokit console artifacts/live/console.log --duration-ms 30000

Boundaries

Device IPs, developer passwords, screenshots, packages, and proof bundles stay local or ignored.
rokit can assert generic Roku and SceneGraph state. Consumer repos own product routes, screen names when product-specific, playback/content expectations, and review narratives.
ECP query paths must be plain paths. Do not include query strings, fragments, traversal, backslashes, or encoded path segments.
The local Effect source setup is skipped whenever CI is set. Do not make CI, install, pack, or publish flows depend on cloning .repos/effect.

rokit

Workflow

Run rokit describe to inspect the command surface instead of guessing flags or --input-json payload fields. Use rokit describe <command> when you only need one command schema.

Use --dry-run before mutating commands such as package, install, launch, press, screenshot, or proof.

Prefer structured output. In automation, rely on the non-TTY JSON default or pass --json explicitly.

Use --fields to keep observations small when only a few values are needed.

Screenshot commands append a timestamp to the requested filename. Use the returned JSON data.path as the artifact path instead of assuming the input path was written.

For live proof, use rokit snapshot for a quick state read, rokit proof <output-dir> for review artifacts, or pnpm live:probe in the rokit repo for the full generic package/install/launch/input/proof probe.

For crash/debug proof, start rokit console <output-path> before reproducing the problem so startup errors and crash traces are captured. Use rokit debug-command <command> only for generic Roku debug commands.

Use rokit wait-ready <app-id> [node-name] [condition] [value] after launch when the app can race ECP or SceneGraph readiness.

Use rokit press <keys...> --until-node <node> --until-state visible for bounded navigation loops instead of arbitrary sleeps.

Start Here

Read only the reference needed for the current task:

live proof, readiness waits, SceneGraph and media-player observation: references/live-proof.md

BrightScript console capture and debug-server commands: references/debugging.md

Common Commands

rokit describe rokit describe proof rokit --json snapshot --fields command,status,data.activeApp,data.mediaPlayer rokit --dry-run package artifacts/live/channel.zip rokit --dry-run launch dev rokit --json proof artifacts/live/proof --screenshot rokit wait-node title text "Ready" rokit --input-json '{"command":"press","keys":["Down","Select"]}' rokit --input-json @artifacts/rokit-payload.json printf '{"command":"press","keys":["Back"]}' | rokit --input-json - rokit console artifacts/live/console.log --duration-ms 30000

Boundaries

Device IPs, developer passwords, screenshots, packages, and proof bundles stay local or ignored.

rokit can assert generic Roku and SceneGraph state. Consumer repos own product routes, screen names when product-specific, playback/content expectations, and review narratives.

ECP query paths must be plain paths. Do not include query strings, fragments, traversal, backslashes, or encoded path segments.

The local Effect source setup is skipped whenever CI is set. Do not make CI, install, pack, or publish flows depend on cloning .repos/effect.