Skip to main content

weekly-demo-video

**WORKFLOW SKILL** — Generate narrated weekly demo videos for azd features. Pulls latest commits, identifies demo-worthy features, researches PRs, and produces MP4 videos with dark-themed slides and neural TTS narration. INVOKES: Python, Pillow, ffmpeg, edge-tts, explore sub-agents, ask_user. USE FOR: weekly demo, generate demo video, demo video, sprint demo, create demo, make demo video, demo for LT, weekly demo video, azd demo. DO NOT USE FOR: general video editing, non-azd demos, slide decks without video, weekly reports (use weekly-report).

Datos de origen

Repositorio
Azure/azure-dev
Última actividad en el origen
20 de mayo de 2026 a las 15:37
Idioma detectado de SKILL.md
inglés
Estrellas
570
Forks
367

Opciones de instalación

De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.

Revisa los archivos de origen

Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.

Explorador de archivos
2 archivos

Mostrando SKILL.md

SKILL.md
Instrucciones de origen · Vista previa de solo lectura
name
weekly-demo-video
license
MIT
metadata
{"version":"1.0"}
description
**WORKFLOW SKILL** — Generate narrated weekly demo videos for azd features. Pulls latest commits, identifies demo-worthy features, researches PRs, and produces MP4 videos with dark-themed slides and neural TTS narration. INVOKES: Python, Pillow, ffmpeg, edge-tts, explore sub-agents, ask_user. USE FOR: weekly demo, generate demo video, demo video, sprint demo, create demo, make demo video, demo for LT, weekly demo video, azd demo. DO NOT USE FOR: general video editing, non-azd demos, slide decks without video, weekly reports (use weekly-report).
# Weekly Demo Video Generator Generates narrated MP4 demo videos for azd features using Python + Pillow + edge-tts + ffmpeg. ## Prerequisites Ensure these tools are installed: | Tool | Purpose | |------|---------| | Python 3 | Script execution | | Pillow | Slide generation (`pip install Pillow`) | | ffmpeg | Video/audio stitch | | edge-tts | Neural TTS (`pip install edge-tts`) | Verify prerequisites (bash): ```bash python3 -c "from PIL import Image; print('ok')" ffmpeg -version edge-tts --list-voices > /dev/null ``` ## Execution Flow ### Step 1: Pull latest ```bash cd <repo-root> # the azure-dev repository root git checkout main && git pull --rebase ``` ### Step 2: Find commits for the week Replace the dates below with the actual week range: ```bash git log --oneline --since="YYYY-MM-DD" --until="YYYY-MM-DD" --no-merges ``` Use the current week window (7 days). For sprint demos, use a 2-week window. ### Step 3: Identify demo-worthy features Group related commits. Skip: typos, CI fixes, test-only, deps bumps, docs-only. Look for: new commands, UX improvements, perf gains, new flags, agent features. Use explore agents in parallel to research each feature group (give them specific commit SHAs). ### Step 4: Confirm with user Present a table of proposed demos. Ask user to confirm or adjust before generating. ### Step 5: Generate videos Follow the conventions strictly: {{ references/CONVENTIONS.md }} ### Step 6: Report List generated videos with filenames and durations. Offer short descriptions for docs. ## Error Handling - **edge-tts failure**: Retry once. If it fails again, log the error and skip that slide's audio — notify the user. - **ffmpeg failure**: Check the ffmpeg error output. Common issues: missing codec, invalid image path. Print the error and stop — don't produce a partial video. - **Font not found**: Falls back to `ImageFont.load_default()` automatically. Warn the user that slides may look different. ## Output All videos go to: `<repo-root>/demo-video/` > **Note:** `demo-video/` is listed in `.gitignore` to prevent accidental commits. ## Demo Naming - **Weekly demos**: `azd_weekly_demo_{date}_{name}.mp4` (e.g. `azd_weekly_demo_may_07_exegraph.mp4`) - **Sprint demos**: `azd_sprint_demo_{date}_{name}.mp4` (e.g. `azd_sprint_demo_apr_28_agent_sessions.mp4`)
Ver en GitHub