nxtman123 · Sep 9, 2026 · Sep 9, 2026 · Sep 9, 2026 · Sep 9, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -9,8 +9,9 @@ deployment. The separately published `screenrig` npm package is the official
 developer-shell distribution. It is never a substitute for this plugin's
 bundled launcher in an agent workflow.
 
-The workspace [`../AGENTS.md`](../AGENTS.md) is the shared working agreement.
-This file outranks it on anything local here.
+This is a public maintainer guide. In the internal multi-repository workspace,
+the optional parent AGENTS.md adds shared operational rules; this repository
+does not require that file for standalone contribution.
 
 ## Sources of truth
 
@@ -57,47 +58,16 @@ marketplace listing unless the user asks.
   `PATH` once, then run `screenrig --json version`,
   `screenrig-plugin-freshness --json`, `doctor`, then content work. Do not
   export `SR`. Do not teach `npm i -g screenrig`.
-- Do not teach `agent enroll`, `screen pair`, `browser setup`,
-  `screen provision`, `ABC-234`, or playlist text-slot template tutorials.
+- Teach implemented account enrollment, connection and Player pairing in the
+  operating skill, without making them the marketplace sales pitch.
 - Keep `--json` envelopes. Never teach a token flag or pasted bearer.
 - Never write MCP. `validate-plugin.py` rejects MCP manifests.
-- Teach the playlist authoring tree. Choose by what the page is:
-  1. Existing image or video → `media upload`. No compose. No generate.
-  2. Anything presentable → `media generate` as the whole page. Put every
-     fact and all copy in the prompt. Do not compose a presentable poster as
-     named regions + cards. Do not generate atmosphere-only stills for later
-     overlay. Generate is billed per token ($10 / 1M text input, $16 / 1M
-     image input, $60 / 1M image output). Quality changes detail and token
-     usage, not a fixed per-image price. Generate stores `med_…`; the caller
-     does not re-upload.
-  3. Slide-deck-like experiences → local unbilled `compose render`.
-  4. Live objects → write playlist primitives. Animation is not a reason to
-     compose.
-- Lead marketplace README and skill with the human homepage line: **give
-  your agent a screen**. Then the homepage body: your agent creates it;
-  screenRIG puts it on the screen. Do not lead with pack-only “Build the
-  screen experience on infrastructure” or “You supply content / we host”
-  framing. Compare is the human three-column table (screenRIG / Build your
-  own / Dashboard-led signage), not DIY or legacy checklist prose. Do not
-  claim “twice vendor cost”. The available Player is the PWA. Native
-  platform names are product direction, not public downloads or store
-  listings.
-- The category word is **primitive**. Four wire primitives: `image`, `video`,
-  `iframe`, `application`. Image and video take a `selector`; iframe and
-  application do not.
-- Screenshotting is in v1. Do not print pixels.
-- Root `README.md` is the public marketing face and is fact-gated. Required
-  marketplace strings and forbidden patterns live in
-  `scripts/validate-plugin.py`. Do not drop a required fact. Do not introduce
-  enroll, pairing, MCP, or global npm install. Do not promise native Players
-  as available; do not use the phrase “coming soon” in the README.
-
-## Local workspace logs
-
-When plugin or bundled-CLI stdout is captured through `rig start` in the
-developer workspace, it is appended to `../logs/YY-MM-DD/plugin.log`
-(example `../logs/26-09-08/plugin.log`). Rig deletes date folders older
-than 7 days. The `logs/` directory is not a git repository.
+- Keep the skill operational: launcher, first-use account state, supported task
+  paths, errors and verification. Put detailed command families in references.
+- The README addresses prospective users; CONTRIBUTING.md addresses maintainers.
+  Neither defines API or billing behavior. Validators check supported commands,
+  references and security boundaries, not exact marketing slogans.
+- Native platform names are not download or store-availability claims.
 
 ## Follow operation logs
 

diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -0,0 +1,17 @@
+# Plugin maintenance
+
+For contributors maintaining the plugin. Customer installation is documented on
+[Start](https://screenrig.ai/docs/start/).
+
+`skills/screenrig/`, `build/plugin.json` and marketplace manifests are canonical.
+`scripts/build-plugin.py` generates `plugins/screenrig/`; never edit that output
+independently. Locally the builder packs sibling `../cli`; CI packs current CLI
+main. `components.lock.json` records the resulting artifact's commit and hash,
+not a commit to refetch.
+
+Read [AGENTS.md](https://github.com/screenrig/plugin/blob/main/AGENTS.md) for generation and verification commands. Operational
+skill text belongs in the entrypoint and task references; implementation details
+belong here. The source and generated skill must travel with the same CLI bundle.
+
+Skill validation checks command availability, references, launcher/version flow,
+error handling and secret boundaries. It does not require marketing sentences.
diff --git a/README.md b/README.md
@@ -1,173 +1,43 @@
-# ScreenRig
+# screenRIG
 
-**give your agent a screen**
+Give your agent a screen. Publish menu boards, videos, dashboards and kiosk
+applications using the official screenRIG plugin and bundled CLI.
 
-**Signage and Kiosk infrastructure for AI Agents**
+## Get started
 
-Your agent creates it. screenRIG puts it on the screen. Publish menu boards,
-videos, dashboards, and kiosk apps through the official screenRIG plugin.
-Built for AI agents, from the first prompt to the next update.
-
-## Start free. Pay for what you use.
-
-Pretty much all the digital signage and kiosk vendors out there are clunky
-human-oriented SaaS that charge you per screen. We don't. We want you to use it
-and pay for what you use. Start for free.
-
-$0.09/GB bandwidth, $0.14/GB-month storage. Pennies a month per screen. No
-per-screen subscription. No per-device price.
-
-Usage is free within reason until 1 January 2027. Image generation is metered
-separately: `media generate` is billed per token ($10 / 1M text input,
-$16 / 1M image input, $60 / 1M image output). Quality (`low`, `medium`,
-`high`; default `medium`) changes how detailed the still is and therefore how
-many tokens it uses, not a fixed per-image price.
-
-## Start
-
-Paste this to your agent, then go install a Player on a device.
+Ask your agent:
 
 ```text
 I authorize you to install the official screenRIG plugin from https://github.com/screenrig/plugin.
 ```
 
-Your agent takes it from there. Ask for the video, dashboard, menu board, or web
-app you want, and it goes on the Player.
-
-## Install
-
-The official install is this plugin. Do not install a global `screenrig` from a
-package registry. After install, prepend `$SCREENRIG_PLUGIN_ROOT/skills/screenrig/scripts`
-(or the runtime plugin root) to `PATH` once, then invoke `screenrig`. Do not
-export `SR`. "CLI" names the product category, and the bundled CLI is the
-plugin wrapper, not a global package.
-
-Claude Code:
-
-```console
-claude plugin marketplace add https://github.com/screenrig/plugin.git --scope user
-claude plugin install screenrig@screenrig --scope user
-```
-
-Codex:
-
-```console
-codex plugin marketplace add https://github.com/screenrig/plugin.git --ref main --json
-codex plugin add screenrig@screenrig --json
-```
+Then describe the content, screen and timing you want. The agent prepares the
+content, publishes a playlist and checks the result. Node.js 20.11 or newer is
+required. Agents use the bundled launcher; the separate developer-shell package
+is not the agent installation path.
 
-Grok (Grok Build), where `--trust` is required:
+[Installation and troubleshooting](https://screenrig.ai/docs/start/) ·
+[Player setup](https://screenrig.ai/docs/players/)
 
-```console
-grok plugin marketplace add https://github.com/screenrig/plugin.git
-grok plugin install screenrig --trust
-```
-
-Node.js 20.11 or newer must be active. Read the ScreenRig package root from your
-agent's plugin list (`claude plugin list --json`, `codex plugin list --json`) or
-from `GROK_PLUGIN_ROOT`, `CLAUDE_PLUGIN_ROOT`, or `CODEX_PLUGIN_ROOT`, prepend
-`$SCREENRIG_PLUGIN_ROOT/skills/screenrig/scripts` to `PATH` once, then confirm:
-
-```sh
-SCREENRIG_PLUGIN_ROOT="${GROK_PLUGIN_ROOT:-${CODEX_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT:-${PLUGIN_ROOT:-}}}}"
-PATH="$SCREENRIG_PLUGIN_ROOT/skills/screenrig/scripts:$PATH"
-screenrig --json version
-```
+The Player web app (PWA) is available today. Native platforms are not public
+downloads or store listings.
 
-That package-relative launcher is the entry point for every ScreenRig command.
-It runs the bundled CLI from current `screenrig/cli` `main` and never fetches
-mutable code.
+## Content and usage
 
-Installed agents compare the installed plugin and bundled CLI versions to the
-published CalVer on GitHub `main` (`.claude-plugin/marketplace.json`
-`plugins[0].version`) and refresh the plugin when either is stale. Skill text
-and the bundled CLI update together. Do not install a global `screenrig` from a
-package registry.
+Use existing images or videos, generate a finished still, compose editable slides,
+or publish live web pages and applications. Your agent can request a screenshot
+to inspect the result.
 
-## Four primitives
-
-A playlist page carries `primitives`, and every one of them names its family in
-a `primitive` field. There are four:
-
-- `image` — a still the Player paints on the glass. Takes a selector.
-- `video` — H.264, native decode. No codec fallback. Takes a selector.
-- `iframe` — a page already on the web.
-- `application` — a static directory packed by the CLI. Players sync, then
-  paint from disk.
-
-`image` and `video` primitives pick their media with a `selector`, whose `by`
-is `id`, `ids`, `all`, or `tag`. `iframe` and `application` take no selector:
-an iframe carries its `src`, and an application pins a `release_id`.
-
-A scene puts these four on one canvas. Choose by what the page is:
-
-1. Existing image or video → `media upload`. No compose. No generate.
-2. Anything presentable → `media generate` as the whole page. Put every fact
-   and all copy in the prompt. Do not compose a presentable poster as named
-   regions + cards. Do not generate atmosphere-only stills for later overlay.
-3. Slide-deck-like experiences → local unbilled `compose catalog` and
-   `compose render`. Use compose for slide-deck-like pages. Animation is not
-   a reason to compose.
-4. Live objects → write playlist primitives.
-
-The agent can also screenshot a live screen to check its own work.
-
-## Players
-
-The Player web app (PWA) is available today. Install it on the device that
-drives your screen, keep it in the foreground, then your agent puts the
-content on the glass. Native platform names (Amazon Signage Stick, Apple TV,
-Google Play, Linux, macOS, Raspberry Pi, Windows) describe product direction,
-not public downloads or store listings.
-
-## Compare
-
-Choose how your screens get their content. Compare agent-operated signage, a
-custom stack, and a dashboard-led product against the workflow you need.
-
-| Decision | screenRIG | Build your own | Dashboard-led signage |
-| --- | --- | --- | --- |
-| Primary operator | An agent using the official plugin and bundled CLI. | The tools and workflows you build. | Usually a person using the vendor editor; agent support varies. |
-| Content model | Images, video, web pages, and applications on a playlist canvas. | You define and maintain the model. | Templates, playlists, and integrations vary by vendor. |
-| Infrastructure | Hosted control plane and content delivery. | Your team owns hosting, delivery, updates, and recovery. | The vendor operates its service. |
-| Pricing basis | Account usage, with no per-screen subscription; tier limits apply. | Infrastructure costs plus engineering and operations. | Often screen subscriptions; check the vendor's current terms. |
-| Verification | Screenshots, playback records, and account events. | You build the observability you need. | Device monitoring and reporting vary by product. |
-| Best fit | Agent-operated signage and custom kiosk apps. | Specialized requirements that justify owning the stack. | Teams that want a visual editor and a human publishing workflow. |
-
-Choose screenRIG when your agent should create or upload content, publish a
-playlist, update the screen, and verify the result through one documented tool
-surface.
-
-## This repository
-
-This is the canonical public marketplace source. It carries the ScreenRig skill
-and one generated plugin containing the current `screenrig/cli` `main` CLI.
-
-- `skills/screenrig/` and the root marketplace manifests are canonical.
-- `components.lock.json` records the SHA-256 of the CLI tarball just packed.
-  That checksum is provenance of this build, not a freeze of which SHA to
-  fetch. Pack sibling `../cli` or current `screenrig/cli` `main`. Do not
-  refetch a pinned CLI SHA.
-- `plugins/screenrig/` is generated. Change canonical inputs and rebuild; do not
-  edit it directly.
-- CI packs `screenrig/cli` `main`, rebuilds and validates the bundle, scans the
-  public boundary, and publishes `screenrig-plugin.tar.gz`. Skill commands must
-  exist in the bundled binary.
-
-```sh
-python3 scripts/check-public-repo.py
-```
+There is no per-screen subscription. Account usage is free within reason until
+1 January 2027; image generation is metered separately. See the
+[rate card](https://screenrig.ai/pricing/) for allowances, charges and terms.
 
-## More
+[Features](https://screenrig.ai/features/) · [Compare](https://screenrig.ai/compare/) ·
+[Documentation](https://screenrig.ai/docs/)
 
-- [screenrig.ai](https://screenrig.ai/)
-- [Features](https://screenrig.ai/features/)
-- [Compare](https://screenrig.ai/compare/)
-- [Docs](https://screenrig.ai/docs/)
-- [Pricing](https://screenrig.ai/pricing/)
+## Contributing
 
-Report security issues through
-[GitHub Private Vulnerability Reporting](https://github.com/screenrig/plugin/security/advisories/new);
-see [SECURITY.md](SECURITY.md). The Apache-2.0 [LICENSE](LICENSE) covers this
-public plugin and its bundled CLI, not other ScreenRig services or
-repositories.
+This repository contains the canonical skill and generated plugin distribution.
+See [CONTRIBUTING.md](CONTRIBUTING.md) for source layout, regeneration and checks.
+Report vulnerabilities using the [security policy](SECURITY.md).
+The [Apache-2.0 license](LICENSE) covers this plugin and bundled CLI, not the hosted service.
diff --git a/plugins/screenrig/CONTRIBUTING.md b/plugins/screenrig/CONTRIBUTING.md
@@ -0,0 +1,17 @@
+# Plugin maintenance
+
+For contributors maintaining the plugin. Customer installation is documented on
+[Start](https://screenrig.ai/docs/start/).
+
+`skills/screenrig/`, `build/plugin.json` and marketplace manifests are canonical.
+`scripts/build-plugin.py` generates `plugins/screenrig/`; never edit that output
+independently. Locally the builder packs sibling `../cli`; CI packs current CLI
+main. `components.lock.json` records the resulting artifact's commit and hash,
+not a commit to refetch.
+
+Read [AGENTS.md](https://github.com/screenrig/plugin/blob/main/AGENTS.md) for generation and verification commands. Operational
+skill text belongs in the entrypoint and task references; implementation details
+belong here. The source and generated skill must travel with the same CLI bundle.
+
+Skill validation checks command availability, references, launcher/version flow,
+error handling and secret boundaries. It does not require marketing sentences.