Video & Audio Production Skills

Inspect Claude Code session JSONL files at ~/.claude/projects/ to extract real conversation telemetry: token counts (input/output/cache reads/cache writes), assistant turn counts, human prompt counts, tool-use counts, compaction boundaries, and the contents of compaction summaries. Use this skill when the user asks "how many tokens did this session use", "how many prompts have I sent", "show me the stats for this conversation", "what got compacted", "where are the compaction boundaries", "introspect the session", "do brain surgery on the JSONL", or wants any data point that lives inside the on-disk session log rather than the live context window. Inspired by Tal Raviv's "I wanted to know how compaction works" article.

Given a local video or video URL, downloads the media if needed, extracts slide frames and key moments, transcribes the audio, and writes a Markdown timeline that interleaves screenshots with the transcript at the associated timestamps. Use when asked to turn a video into a multimodal notes file, slide-synced transcript, screenshot-enhanced transcript, or talk recap with images.

Fully automated new Mac setup for fullstack web developers and AI engineers. Generates and executes idempotent shell scripts to install dev tools, CLI utilities, GUI apps, AI/ML tooling, shell config, and macOS preferences — all via Homebrew, fnm, uv, and defaults commands. Use this skill whenever someone says "set up my Mac", "new Mac setup", "configure my dev machine", "install my dev tools", "fresh Mac install", "developer environment setup", or anything about bootstrapping a macOS machine for development. Also trigger when someone mentions setting up Homebrew, Oh-My-ZSH, Node.js, Python, Docker, or terminal configuration on a new Mac — even if they don't call it a "setup". This skill is opinionated and based on real-world experience setting up M-series Macs for fullstack JS/TS + Python + AI development.

Transcribe long-form audio, YouTube videos, podcasts, interviews, or panels; summarize them; extract chapter markers; and draft publishing assets like titles, YouTube descriptions, show notes, and tweet/X copy. Use when a user shares a YouTube link or audio/video file and asks for transcription, diarization, summaries with timestamps, chapter markers, show notes, titles, descriptions, promo copy, or social posts. Especially use for Whisper-based workflows, source-audio-first podcast processing, and any request to turn a recorded conversation into publishable assets.

Summarizes arbitrarily long text (1k-1M words) using recursive map-reduce with any LLM backend. Accepts raw text, markdown, transcripts, articles, codebases, or any plaintext input. Produces one or more output formats: executive summary, section headings with timestamps, YouTube description, Twitter/X posts, title options, thumbnail prompts, blog outlines, pull quotes, and more. Supports focus directives ("focus on the AI parts", "emphasize the business angle") to steer the summary. Pluggable backends: OpenRouter, Ollama, OpenAI, Anthropic, Gemini, or any OpenAI-compatible endpoint. Use this skill when someone says "summarize this", "give me a summary", "TL;DR", "make this shorter", "create a YouTube description", "write a tweet about this", "generate titles", "thumbnail ideas", or provides long text and wants any condensed output.

Extracts the most interesting frames from video files for thumbnail compositing. Detects faces, expressions, smiles, and presentation slides. Outputs full frames, face crops, and transparent cutouts. Use when asked to extract thumbnails, find interesting frames, grab screenshots from video, or create thumbnail candidates from recordings.

Transcribes audio and video files to text using pluggable ASR backends. Default backend is local whisper CLI (openai-whisper). Supports whisperX (with diarization), insanely-fast-whisper, faster-whisper, whisper.cpp, OpenAI Whisper API, Groq Whisper API, Deepgram, AssemblyAI, Gemini, and Hugging Face models. Handles very long files (1-8+ hours) by preprocessing with ffmpeg: extracts audio from video, converts to optimal ASR format, detects and skips silence, and chunks for API size limits. Supports speaker diarization, word-level timestamps, custom vocabulary, and multiple output formats. Use this skill when someone says "transcribe this", "convert to text", "speech to text", "get the transcript", "transcribe this video/audio/podcast/recording", or provides a media file and wants text output.

Use when the user asks to "create YouTube thumbnails", "generate thumbnails", "make video thumbnails", or needs AI-generated thumbnail images via Google Gemini for YouTube videos. Covers prompt engineering, image generation, compression, and upload to YouTube Studio.

Use when the user asks to "download Zoom recordings", "grab recordings from Zoom", "get new Zoom videos", or needs to download cloud recordings and analyze their content via frame extraction. Covers file type selection, filename verification, and ffmpeg-based content analysis.

Use when the user asks to "upload Zoom recordings to YouTube", "publish Zoom videos", "download and upload recordings", or wants to run the full Zoom-to-YouTube pipeline. It orchestrates the zoom-download, youtube-publish, and youtube-thumbnails sub-skills with checkpoint stages between each phase.