@growthub/cli 0.3.54 → 0.3.55

package/assets/worker-kits/growthub-open-montage-studio-v1/brands/growthub/brand-kit.md

@@ -0,0 +1,50 @@
+# Brand Kit — Growthub
+
+## Brand Identity
+
+| Field | Value |
+|---|---|
+| Brand name | Growthub |
+| Tagline | Growth infrastructure over a stable agentic substrate |
+| Industry | AI / Developer tools / Growth automation |
+| Primary audience | Developers, growth teams, AI-native companies |
+
+## Visual Identity
+
+| Field | Value |
+|---|---|
+| Primary color | `#1a1a2e` (deep navy) |
+| Secondary color | `#16213e` (dark blue) |
+| Accent color | `#0f3460` (medium blue) |
+| Background | `#0a0a0a` (near black) |
+| Font family (headings) | Monospace / terminal aesthetic |
+| Font family (body) | System sans-serif |
+| Logo path | ASCII art block letter style |
+
+## Tone of Voice
+
+Technical, direct, confident. No marketing fluff. Speak to builders and operators who value substance over polish. Use declarative statements. Avoid exclamation marks.
+
+## Visual Style Preferences
+
+- Clean professional or minimalist diagram style for educational content
+- Flat motion graphics for social media
+- Dark backgrounds with high-contrast text and subtle accent colors
+- Terminal/code aesthetic where appropriate
+- Avoid: stock photo look, corporate clip art, generic gradients
+
+## Platform Profiles
+
+| Platform | Aspect Ratio | Duration Range | Notes |
+|---|---|---|---|
+| YouTube | 16:9 | 30-300s | Primary distribution |
+| TikTok | 9:16 | 15-60s | Short demos |
+| LinkedIn | 16:9 | 30-120s | Product announcements |
+
+## Music Preferences
+
+Ambient electronic, minimal beats, lo-fi. Instrumental only. Low energy for educational, medium energy for product demos. No vocals.
+
+## Deliverable Log
+
+<!-- Append deliverables as they are produced. -->

package/assets/worker-kits/growthub-open-montage-studio-v1/bundles/growthub-open-montage-studio-v1.json

@@ -0,0 +1,55 @@
+{
+  "schemaVersion": 2,
+  "bundle": {
+    "id": "growthub-open-montage-studio-v1",
+    "version": "1.0.0",
+    "kitId": "growthub-open-montage-studio-v1",
+    "workerId": "open-montage-studio-operator"
+  },
+  "briefType": "open-montage-video-production",
+  "publicExampleBrandPaths": [
+    "brands/growthub/brand-kit.md"
+  ],
+  "requiredFrozenAssets": [
+    "QUICKSTART.md",
+    ".env.example",
+    "skills.md",
+    "output-standards.md",
+    "runtime-assumptions.md",
+    "validation-checklist.md",
+    "workers/open-montage-studio-operator/CLAUDE.md",
+    "brands/_template/brand-kit.md",
+    "brands/growthub/brand-kit.md",
+    "brands/NEW-CLIENT.md",
+    "setup/clone-fork.sh",
+    "setup/verify-env.mjs",
+    "setup/check-deps.sh",
+    "output/README.md",
+    "templates/video-production-brief.md",
+    "templates/pipeline-selection-brief.md",
+    "templates/provider-selection-brief.md",
+    "templates/scene-plan.md",
+    "templates/prompt-matrix.md",
+    "templates/generation-batch-plan.md",
+    "templates/asset-tracking.md",
+    "templates/review-qa-checklist.md",
+    "templates/platform-ready-execution-handoff.md",
+    "templates/cms-node-pipeline-mapping.md",
+    "examples/video-production-brief-sample.md",
+    "examples/pipeline-selection-sample.md",
+    "examples/cms-node-video-gen-sample.md",
+    "examples/platform-ready-handoff-sample.md",
+    "docs/open-montage-fork-integration.md",
+    "docs/provider-adapter-layer.md",
+    "docs/cms-node-bridge.md",
+    "docs/pipeline-reference.md",
+    "growthub-meta/README.md",
+    "growthub-meta/kit-standard.md"
+  ],
+  "optionalPresets": [],
+  "export": {
+    "folderName": "growthub-agent-worker-kit-open-montage-studio-v1",
+    "zipFileName": "growthub-agent-worker-kit-open-montage-studio-v1.zip"
+  },
+  "activationModes": ["export"]
+}

package/assets/worker-kits/growthub-open-montage-studio-v1/docs/cms-node-bridge.md

@@ -0,0 +1,152 @@
+# CMS Node Bridge — GrowthHub ↔ OpenMontage
+
+This document describes the adapter layer that bridges GrowthHub CMS node outputs into the OpenMontage production pipeline.
+
+---
+
+## Overview
+
+GrowthHub's CMS exposes video and image generation as hosted execution nodes. These nodes produce output URLs (video clips, images, audio) that can feed directly into OpenMontage's composition engine. The bridge is a **cognitive adapter** — the agent reads this mapping and applies it during pipeline execution. There is no code-level integration; the agent is the integration layer.
+
+---
+
+## Architecture
+
+```
+GrowthHub CMS                          OpenMontage
+─────────────                          ───────────
+                                       
+CMS Capability Registry ──→ Agent discovers available nodes
+       │                              
+       ↓                              
+CMS Node Execution ──→ Agent triggers hosted execution
+       │                              
+       ↓                              
+Output URLs ──→ Agent feeds URLs as source assets
+       │                     │
+       ↓                     ↓
+                    OpenMontage Pipeline Stage
+                    (assets / edit / compose)
+                             │
+                             ↓
+                    Remotion / FFmpeg Composition
+                             │
+                             ↓
+                    Final Video Output
+```
+
+---
+
+## CMS Capability Registry
+
+The CMS capability registry (`cli/src/runtime/cms-capability-registry/`) exposes node primitives:
+
+```typescript
+interface CmsCapabilityNode {
+  slug: string;                    // e.g., "video-generation"
+  family: CapabilityFamily;        // "video" | "image" | "slides" | "text" | "data" | "ops"
+  executionKind: string;           // "hosted-execute"
+  executionBinding: {
+    type: "mcp_tool_call";
+    strategy: "direct" | "sequential-with-persistence" | "async_operation";
+  };
+  requiredBindings: string[];      // Provider keys needed
+  outputTypes: string[];           // e.g., ["video_url", "image_url"]
+  enabled: boolean;
+}
+```
+
+The agent queries the registry to discover which nodes are available:
+
+```
+Query: { family: "video", enabledOnly: true }
+Query: { family: "image", enabledOnly: true }
+```
+
+---
+
+## Node Family → OpenMontage Stage Mapping
+
+| CMS Family | Output Type | OpenMontage Stage | Integration |
+|---|---|---|---|
+| `video` | Video URL | `assets` | Download → source clip for composition |
+| `image` | Image URL | `assets` | Download → scene image for Remotion animation |
+| `text` | String | `script` | Direct use as script content |
+| `slides` | Slide assets | `assets` | Scene images extracted from slides |
+
+---
+
+## Execution Flow
+
+### 1. Discovery
+
+Agent queries the CMS capability registry during Step 6 (provider selection):
+
+```
+Available CMS nodes:
+- video-generation (family: video, hosted-execute)
+- image-generation (family: image, hosted-execute)
+```
+
+### 2. Input Binding Preparation
+
+For each scene that will use a CMS node, the agent prepares input bindings from the scene plan:
+
+```
+CMS Node: image-generation
+Inputs: { prompt: "...", aspect_ratio: "16:9", style: "photorealistic" }
+```
+
+### 3. Hosted Execution
+
+The agent triggers execution via the hosted execution client. Results include output URLs:
+
+```
+Result: { status: "completed", outputUrl: "https://cdn.growthub.ai/gen/abc123.png" }
+```
+
+### 4. Asset Integration
+
+Output URLs are recorded in the asset tracking sheet and fed into the OpenMontage pipeline:
+
+- **Video URLs** → downloaded as source clips → timeline assembly via FFmpeg or Remotion
+- **Image URLs** → downloaded as scene images → animated composition via Remotion
+
+### 5. Composition
+
+OpenMontage handles all post-production:
+- Scene ordering and timeline assembly
+- Transition effects (crossfade, zoom, wipe)
+- Audio mixing (narration + music)
+- Subtitle burn-in (word-level timing)
+- Color grading
+- Final render to target platform spec
+
+---
+
+## When to Bridge vs When to Use Direct Providers
+
+| Scenario | Recommendation |
+|---|---|
+| User has GrowthHub session, no direct API keys | Bridge via CMS nodes |
+| User prefers unified GrowthHub billing | Bridge via CMS nodes |
+| CMS node matches the required generation type | Bridge via CMS nodes |
+| User has direct API keys and wants lower latency | Use direct providers |
+| Required tool/model is not available as a CMS node | Use direct providers |
+| Local-fork mode with full tool access | Use direct providers (faster) |
+
+---
+
+## Adapter Contract
+
+The bridge adapter handles:
+
+| Operation | Description |
+|---|---|
+| `DISCOVER_NODES` | Query CMS capability registry for available nodes |
+| `PREPARE_BINDINGS` | Map scene plan inputs to CMS node input schema |
+| `EXECUTE_NODE` | Trigger hosted execution and poll for results |
+| `COLLECT_OUTPUTS` | Gather output URLs from execution results |
+| `ROUTE_TO_PIPELINE` | Feed output URLs into the appropriate OpenMontage pipeline stage |
+
+The agent performs all of these operations as part of the standard 10-step workflow. No separate adapter code is needed — the agent IS the adapter.

package/assets/worker-kits/growthub-open-montage-studio-v1/docs/open-montage-fork-integration.md

@@ -0,0 +1,120 @@
+# OpenMontage Fork Integration
+
+This document describes how the local OpenMontage clone integrates with this worker kit.
+
+---
+
+## Architecture Overview
+
+OpenMontage is an agent-first video production system. There is no code orchestrator — the AI agent IS the orchestrator.
+
+```
+Agent reads pipeline manifest (YAML) → stages, tools, review criteria
+  ↓
+Agent reads stage director skill (Markdown) → HOW to execute each stage
+  ↓
+Agent calls Python tools → scored provider selection across 7 dimensions
+  ↓
+Agent self-reviews → schema validation, quality checks
+  ↓
+Agent checkpoints state (JSON) → resumable, with decision log
+  ↓
+Pre-compose validation gate → delivery promise, slideshow risk
+  ↓
+Render (Remotion or FFmpeg)
+  ↓
+Post-render self-review → ffprobe, frame extraction, audio analysis
+  ↓
+Final video output
+```
+
+---
+
+## Directory Layout
+
+```
+OpenMontage/
+├── tools/              # 52 Python tools (the agent's hands)
+│   ├── video/          # Video gen: Kling, Runway, Veo, Grok, WAN, Hunyuan, etc.
+│   ├── audio/          # TTS: ElevenLabs, Google, OpenAI, Piper + music
+│   ├── graphics/       # Image gen: FLUX, DALL-E, Imagen, Recraft, local diffusion
+│   ├── enhancement/    # Upscale, bg remove, face enhance, color grade
+│   ├── analysis/       # Transcription, scene detect, frame sampling
+│   ├── avatar/         # Talking head, lip sync
+│   └── subtitle/       # SRT/VTT generation
+│
+├── pipeline_defs/      # 12 YAML pipeline manifests
+├── skills/             # 400+ Markdown skill files
+│   ├── pipelines/      # Per-pipeline stage director skills
+│   ├── creative/       # Creative technique skills
+│   ├── core/           # Core tool skills
+│   └── meta/           # Reviewer, checkpoint protocol
+│
+├── schemas/            # JSON Schemas for contract validation
+├── styles/             # Visual style playbooks (YAML)
+├── remotion-composer/  # React/Remotion video composition engine
+├── lib/                # Core infrastructure
+└── config.yaml         # Global configuration
+```
+
+---
+
+## Tool Discovery
+
+The tool registry auto-discovers all tools at startup:
+
+```bash
+cd ~/OpenMontage
+python3 -c "from tools.tool_registry import registry; import json; registry.discover(); print(json.dumps(registry.support_envelope(), indent=2))"
+```
+
+This returns the full capability envelope: which tools are available, which providers are configured, and what the system can produce.
+
+---
+
+## Three-Layer Knowledge Architecture
+
+| Layer | Location | Purpose |
+|---|---|---|
+| 1 | `tools/` + `pipeline_defs/` | What exists — executable capabilities + orchestration |
+| 2 | `skills/` | How to use it — OpenMontage conventions and quality bars |
+| 3 | `.agents/skills/` | How it works — 47 external technology knowledge packs |
+
+---
+
+## Pipeline Manifests
+
+Each pipeline is defined as a YAML manifest in `pipeline_defs/`. The manifest specifies:
+- Stages in order
+- Tools available at each stage
+- Review criteria
+- Success gates
+
+The agent reads the manifest, then reads the corresponding stage director skills to know HOW to execute each stage.
+
+---
+
+## Remotion Composition Engine
+
+Located at `remotion-composer/`. A React-based programmatic video composition engine that:
+- Renders TextCard, StatCard, charts, image scenes with spring animations
+- Supports crossfade, wipe, zoom transitions
+- Burns in TikTok-style word-level captions
+- Outputs H.264 MP4 at any resolution/aspect ratio
+
+Remotion is always available (no API key needed) and is the primary composition path for image-based video.
+
+---
+
+## Setup
+
+```bash
+bash setup/clone-fork.sh
+```
+
+This script:
+1. Clones `https://github.com/calesthio/OpenMontage.git` to `~/OpenMontage`
+2. Runs `make setup` (pip install + npm install + Piper TTS)
+3. Verifies the tool registry
+
+After setup, add API keys to `~/OpenMontage/.env` for additional provider access.

package/assets/worker-kits/growthub-open-montage-studio-v1/docs/pipeline-reference.md

@@ -0,0 +1,147 @@
+# OpenMontage Pipeline Reference
+
+Quick reference for all 12 OpenMontage production pipelines.
+
+---
+
+## Pipelines
+
+### Animated Explainer
+
+| Field | Value |
+|---|---|
+| Best for | Educational content, tutorials, topic breakdowns |
+| Stage flow | research → proposal → script → scene_plan → assets → edit → compose |
+| Key tools | Web search, image gen (FLUX/DALL-E), TTS, Remotion, FFmpeg |
+| Typical cost | $0.15-$3.00 |
+| Duration range | 30-300 seconds |
+
+### Animation
+
+| Field | Value |
+|---|---|
+| Best for | Social media, product demos, abstract concepts |
+| Stage flow | research → proposal → script → scene_plan → assets → edit → compose |
+| Key tools | Image gen, video gen (optional), Remotion motion graphics |
+| Typical cost | $0.15-$5.00 |
+| Duration range | 15-120 seconds |
+
+### Avatar Spokesperson
+
+| Field | Value |
+|---|---|
+| Best for | Corporate comms, training, announcements |
+| Stage flow | research → proposal → script → scene_plan → assets → edit → compose |
+| Key tools | HeyGen avatar, TTS, Remotion |
+| Typical cost | $2.00-$10.00 |
+| Duration range | 30-300 seconds |
+
+### Cinematic
+
+| Field | Value |
+|---|---|
+| Best for | Brand films, teasers, promotional content |
+| Stage flow | research → proposal → script → scene_plan → assets → edit → compose |
+| Key tools | Video gen (Kling/Veo/Runway), image gen, music gen, FFmpeg |
+| Typical cost | $1.00-$10.00 |
+| Duration range | 15-120 seconds |
+
+### Clip Factory
+
+| Field | Value |
+|---|---|
+| Best for | Repurposing long content for social media |
+| Stage flow | analysis → ranking → extraction → edit → compose |
+| Key tools | Transcriber, scene detect, FFmpeg trim, subtitle gen |
+| Typical cost | $0.00-$1.00 |
+| Duration range | 15-60 seconds per clip |
+
+### Documentary Montage
+
+| Field | Value |
+|---|---|
+| Best for | Video essays, mood pieces, real-footage B-roll edits |
+| Stage flow | research → corpus_build → clip_selection → edit → compose |
+| Key tools | Pexels/Archive.org/Wikimedia, CLIP search, FFmpeg, music |
+| Typical cost | $0.00-$2.00 |
+| Duration range | 30-300 seconds |
+
+### Hybrid
+
+| Field | Value |
+|---|---|
+| Best for | Enhancing existing footage with AI-generated support visuals |
+| Stage flow | analysis → proposal → assets → edit → compose |
+| Key tools | Scene detect, image gen, video gen, FFmpeg overlay |
+| Typical cost | $0.50-$5.00 |
+| Duration range | 15-300 seconds |
+
+### Localization & Dub
+
+| Field | Value |
+|---|---|
+| Best for | Multi-language distribution |
+| Stage flow | analysis → transcription → translation → dub → compose |
+| Key tools | Transcriber, TTS (multi-language), subtitle gen, FFmpeg |
+| Typical cost | $0.50-$5.00 per language |
+| Duration range | Matches source |
+
+### Podcast Repurpose
+
+| Field | Value |
+|---|---|
+| Best for | Podcast marketing, audiogram videos |
+| Stage flow | analysis → highlight_extraction → visual_design → compose |
+| Key tools | Transcriber, image gen, Remotion audiogram, FFmpeg |
+| Typical cost | $0.00-$2.00 |
+| Duration range | 30-120 seconds per clip |
+
+### Screen Demo
+
+| Field | Value |
+|---|---|
+| Best for | Product demos, tutorials, documentation |
+| Stage flow | proposal → script → recording → edit → compose |
+| Key tools | Screen capture, TTS narration, Remotion overlays, FFmpeg |
+| Typical cost | $0.00-$1.00 |
+| Duration range | 30-600 seconds |
+
+### Talking Head
+
+| Field | Value |
+|---|---|
+| Best for | Presentations, vlogs, interviews |
+| Stage flow | proposal → script → recording → edit → compose |
+| Key tools | Source footage, face enhance, TTS (optional), FFmpeg |
+| Typical cost | $0.00-$2.00 |
+| Duration range | 30-600 seconds |
+
+---
+
+## Universal Stage Flow
+
+All pipelines follow the same core structure:
+
+```
+research → proposal → script → scene_plan → assets → edit → compose
+```
+
+Some pipelines have specialized stages (e.g., `corpus_build` for Documentary Montage, `analysis` for Clip Factory), but the overall flow is always: understand → plan → generate → assemble → deliver.
+
+---
+
+## Pipeline Selection Decision Matrix
+
+| If the user wants... | Select... |
+|---|---|
+| Explain a topic | Animated Explainer |
+| Motion graphics / kinetic typography | Animation |
+| Avatar presenter | Avatar Spokesperson |
+| Cinematic trailer / brand film | Cinematic |
+| Batch clips from long content | Clip Factory |
+| Real footage montage / video essay | Documentary Montage |
+| Enhance existing footage | Hybrid |
+| Translate / dub | Localization & Dub |
+| Podcast to video | Podcast Repurpose |
+| Software walkthrough | Screen Demo |
+| Speaker-led video | Talking Head |

package/assets/worker-kits/growthub-open-montage-studio-v1/docs/provider-adapter-layer.md

@@ -0,0 +1,105 @@
+# Provider Adapter Layer
+
+OpenMontage uses a scored provider selection system. This document describes how providers integrate and how GrowthHub CMS nodes fit into the provider model.
+
+---
+
+## Provider Selection Model
+
+Every tool selection runs through a 7-dimension scoring engine:
+
+| Dimension | Weight | Description |
+|---|---|---|
+| Task fit | 30% | How well the provider matches the specific task |
+| Output quality | 20% | Resolution, fidelity, consistency |
+| Control features | 15% | Prompt adherence, parameter granularity |
+| Reliability | 15% | Uptime, error rate, queue latency |
+| Cost efficiency | 10% | Price per output unit |
+| Latency | 5% | Request to deliverable time |
+| Continuity | 5% | Session persistence, rate limits |
+
+The agent scores available providers for each task and selects the highest-scoring option. Alternatives are logged with their scores in the decision audit trail.
+
+---
+
+## Provider Categories
+
+### Cloud API Providers
+
+| Provider | Env Key | Capabilities |
+|---|---|---|
+| fal.ai | `FAL_KEY` | FLUX images, Kling/Veo/MiniMax video, Recraft images |
+| OpenAI | `OPENAI_API_KEY` | DALL-E 3 images, OpenAI TTS |
+| Google | `GOOGLE_API_KEY` | Google TTS (700+ voices), Imagen images |
+| ElevenLabs | `ELEVENLABS_API_KEY` | Premium TTS, music, SFX |
+| xAI Grok | `XAI_API_KEY` | Image + video generation |
+| Runway | `RUNWAY_API_KEY` | Gen-4 video |
+| Higgsfield | `HIGGSFIELD_API_KEY` | Multi-model video orchestrator |
+| HeyGen | `HEYGEN_API_KEY` | Avatar videos |
+| Suno | `SUNO_API_KEY` | AI music generation |
+
+### Free Stock Media
+
+| Provider | Env Key | Capabilities |
+|---|---|---|
+| Pexels | `PEXELS_API_KEY` | Stock photos + videos |
+| Pixabay | `PIXABAY_API_KEY` | Stock photos + videos |
+| Unsplash | `UNSPLASH_ACCESS_KEY` | Stock images |
+
+### Local Providers (Always Available)
+
+| Provider | Capabilities |
+|---|---|
+| Piper TTS | Offline text-to-speech, ~30 English voices |
+| Remotion | React-based programmatic video composition |
+| FFmpeg | Video encoding, subtitles, audio mixing, color grading |
+| Archive.org + NASA + Wikimedia | Free archival footage |
+
+### Local GPU Providers (Optional)
+
+| Provider | Env Key | Requirements |
+|---|---|---|
+| WAN 2.1 | `VIDEO_GEN_LOCAL_ENABLED` | NVIDIA GPU, 6-24GB VRAM |
+| Hunyuan | `VIDEO_GEN_LOCAL_ENABLED` | NVIDIA GPU, 12GB VRAM |
+| CogVideo | `VIDEO_GEN_LOCAL_ENABLED` | NVIDIA GPU, 6-10GB VRAM |
+| LTX-Video | `VIDEO_GEN_LOCAL_ENABLED` | NVIDIA GPU, 8GB VRAM |
+| Stable Diffusion | — | NVIDIA GPU, 4GB+ VRAM |
+
+---
+
+## GrowthHub CMS Nodes as Providers
+
+GrowthHub CMS nodes act as an additional provider source in the scoring model. They are treated identically to cloud providers with the following characteristics:
+
+| Dimension | CMS Node Score Basis |
+|---|---|
+| Task fit | Matches if the CMS node family matches the task (video/image) |
+| Output quality | Depends on the underlying provider routed by the CMS node |
+| Control features | Limited to CMS node input bindings |
+| Reliability | GrowthHub hosted infrastructure SLA |
+| Cost efficiency | GrowthHub billing (may be included in plan) |
+| Latency | Hosted execution overhead + provider latency |
+| Continuity | Depends on GrowthHub session |
+
+When CMS nodes are available and the user prefers unified GrowthHub billing, the scoring engine should weight cost efficiency higher for CMS nodes.
+
+---
+
+## Fallback Strategy
+
+When the primary provider path fails:
+
+1. Fall back to the next highest-scored provider for the same task
+2. If no alternative provider exists, fall back to a different tool category (e.g., image-to-video instead of text-to-video)
+3. If all providers fail, fall back to free/local tools (Piper TTS, Remotion composition, stock media)
+4. Mark the handoff with the exact fallback trigger and trade-offs
+
+---
+
+## Adding a New Provider
+
+New providers integrate by:
+1. Adding a tool in the appropriate `tools/` subdirectory
+2. Implementing the tool contract (BaseTool inheritance)
+3. The registry auto-discovers it
+4. The scoring engine includes it in selection automatically