Maagpi Youtube Workflow Skill
About
Automate end-to-end youtube production pipeline for face less channels
Details
- Author
- vamsi-kodimela
- Categories
- Other, Automation, Media
Jump to
Setup
Install Maagpi Youtube Workflow Skill in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/vamsi-kodimela/maagpi-youtube-workflow-skill
Follow the installation instructions in the repository README, then restart your MCP client.
Automate end-to-end youtube production pipeline for face less channels
YouTube Content Workflow is aClaude Code skillthat turns a Notion content calendar into scheduled YouTube videos end-to-end: pick a channel, generate SUCCESS-framework titles, run NotebookLMdeep research, generate an Explainer video in the channel's language, transcribe it, draft the description, build the thumbnail (preferring Nano Banana Pro), generate tags, and schedule upload —never public.
This repository is the source for the skill. After install it lives at~/.claude/skills/youtube-content-workflow/and is invoked via the/youtube-content-workflowslash command.
The fastest path from zero to a scheduled upload:
See thePhase Referencefor what happens at each step and theExamplefor a full annotated run.
- Trigger:/youtube-content-workflow
- Install path:~/.claude/skills/youtube-content-workflow/
- State:local JSON cache (per-channel + per-run) + Notion (canonical)
- Full docs:seeSKILL.mdfor the complete phase-by-phase contract
The skill orchestrates external MCPs at runtime.Install these in your Claude environment before first use— Phase 0 of the skill probes each one and halts with the exact missing-tool name if any are absent.
- Notion MCP— read the content calendar, write per-channel context
- NotebookLM MCP—notebook_create,research_start(deep mode),studio_create,download_artifact· install:https://mcpservers.org/servers/roomi-fields/notebooklm-mcp
- YouTube MCP— list user's channels, fetch recent video titles, schedule uploads with privacyprivate/unlisted· install:https://github.com/vamsi-kodimela/maagpi-youtube-mcp
- Image generation MCP— Nano Banana Pro preferred; falls back to other Gemini/Imagen tools at runtime · install:https://github.com/vamsi-kodimela/maagpi-images-mcp
- Transcription— dedicated MCP, NotebookLMstudio_statustranscript field, or localwhisper(auto-detected, in that order)
Don't have these MCPs yet? The links above point to ready-to-install servers. Phase 0.1 prints them in the halt message if anything is missing.
Never publish public. The skill enforcesprivacy != publicwith a pre-flight assertion before every upload call. There is no opt-out.
If the package isn't on npm yet, install straight from GitHub:
npx github:vamsi-kodimela/maagpi-youtube-workflow-skill install
npx youtube-content-workflow install # idempotent; safe to re-run npx youtube-content-workflow uninstall # KEEPS state/ npx youtube-content-workflow uninstall --purge # also delete state/ (irreversible) npx youtube-content-workflow uninstall --yes # skip confirmation npx youtube-content-workflow --help
Requires Node ≥ 18. The CLI is cross-platform (macOS, Linux, Windows) and does the same work as the shell installers below.
git clone https://github.com/vamsi-kodimela/maagpi-youtube-workflow-skill.git cd maagpi-youtube-workflow-skill chmod +x install.sh ./install.sh
git clone https://github.com/vamsi-kodimela/maagpi-youtube-workflow-skill.git cd maagpi-youtube-workflow-skill Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass -Force ./install.ps1
cd maagpi-youtube-workflow-skill ./install.sh
mkdir -p ~/.claude/skills/youtube-content-workflow/{schemas,state/channels,state/runs} cp SKILL.md ~/.claude/skills/youtube-content-workflow/ cp schemas/ ~/.claude/skills/youtube-content-workflow/schemas/
Then append the registration block to~/.claude/CLAUDE.mdso the slash command dispatches:
# youtube-content-workflow - youtube-content-workflow (~/.claude/skills/youtube-content-workflow/SKILL.md) - End-to-end YouTube production pipeline. Trigger: /youtube-content-workflow When the user types /youtube-content-workflow, invoke the Skill tool with skill: "youtube-content-workflow" before doing anything else.
All installers are idempotent — re-run after editingSKILL.mdto push the new copy.
The skill runs five phases. Phases 2 and 3 fan out as parallelAgentsubagents (one per topic). User gates batch sequentially.
A typical run for three topics in the next 7 days:
> /youtube-content-workflow [Phase 0] channel = Main Channel (UCxxxxxxxx) — cache hit [Phase 1] 3 topics found in date range 2026-04-30..2026-05-07 [Phase 2] spawned 3 parallel title gens topic 1: 5 candidates → user picked #2 topic 2: 5 candidates → user picked #4 topic 3: 5 candidates → user picked #1 [Phase 3] spawned 3 parallel pipelines topic 1: notebook ✓ deep research (8m 12s) ✓ video (3m 4s) ✓ transcript ✓ description ✓ thumbnail ✓ tags ✓ topic 2: ... topic 3: ... [Phase 4] review topic 1: approved review topic 2: regen thumbnail → approved review topic 3: approved privacy: private [Phase 5] pre-flight pass · uploaded 3 · all scheduled as private ✅ run complete: ~/.claude/skills/youtube-content-workflow/state/runs/run-2026-04-30-a1b2c3/
~/.claude/skills/youtube-content-workflow/ ├── SKILL.md ├── schemas/ └── state/ ├── channels/ │ └── UCxxxxxxxx.json # cached channel metadata (mirrors Notion) └── runs/ └── run-2026-04-30-a1b2c3/ └── <topic_slug>/ ├── metadata.json ├── video.mp4 ├── thumbnail.png ├── transcript.txt ├── description.txt ├── tags.json └── titles.json
Channel state iscanonical in Notionand locally cached for speed. Seeschemas/notion-databases.mdfor the required Notion schema (YouTube Channels+Content Calendar).
EditSKILL.mdin this repo, then re-run any installer (npx youtube-content-workflow install,./install.sh, or./install.ps1). The installers overwrite the deployed copy and skip the CLAUDE.md append if the registration block is already present.
npx youtube-content-workflow uninstall # KEEPS state/ npx youtube-content-workflow uninstall --purge # also delete state/ (irreversible) npx youtube-content-workflow uninstall --yes # skip confirmation
./uninstall.sh # remove SKILL.md + schemas + CLAUDE.md block; KEEP state/ ./uninstall.sh --purge # also delete state/ (irreversible) ./uninstall.sh --yes # skip confirmation
./uninstall.ps1 # preserves state/ ./uninstall.ps1 -Purge # also delete state/ ./uninstall.ps1 -Yes # skip confirmation
- Anti-hallucination by construction.Tool-sourced facts only, per-step output verification, language preservation. Never invents topics, calendar entries, or transcripts.
- Parallel by default.Title gen and the heavy per-topic pipeline both fan out as subagent teams; user gates (title pick, review, privacy) batch sequentially.
- Never public.Default privacy isprivate.unlistedrequires explicit opt-in. A pre-flight assertion enforces this before every upload call.
- Skill body:SKILL.md
- Notion DB schemas:schemas/notion-databases.md
- Channel cache schema:schemas/channel-state.example.json
- Reference: SUCCESS framework — Heath brothers,Made to Stick*
The legacyyoutube-content-creatorandyoutube-notebooklmskills at~/.claude/skills/are left intact. This skill is a unified, MCP-first replacement. To retire the old ones, delete their directories and remove their CLAUDE.md entries manually — neither installer touches them.
Local MCP server for YouTube metadata, thumbnails, comments, and analytics.
All-in-one AI creative studio — generate videos, images, audio in 11 Indian languages, and 3D models via MCP. Hosted at mcp.arcframe.ai.
AI podcast editing as a service. Upload raw audio or submit a URL, get back edited episodes with filler words removed, noise reduction, transcripts, show notes, and social clips. Includes webhooks for automation.
Official MCP server for Comfy Cloud - generate images, video, audio, and 3D, run ComfyUI workflows, and search models directly from Claude and other AI agents
34-tool MCP server for CutPro's v1 API — analyze videos, clip, render, and publish posts via AI clients like Claude, ChatGPT, and Cursor.
A server for creating fast and free lipsync videos for digital avatars, supporting both realistic and cartoon styles.
https://github.com/vamsi-kodimela/maagpi-youtube-mcp
Licensed AI content access. 1,821 Dhar Mann Studios videos. $0.0025/pull, creator compensated automatically. 30ms delivery.
AI photo generation, editing, and video creation. 150+ themes, 19 edit operations, 3 video types.
Generate, edit, and render videos, images, and audio from Claude, Cursor, ChatGPT and other AI tools via the Shotstack video editing API.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
