summarizelisted

# tubeless — YouTube video summaries Take a YouTube URL and run **transcript -> summary**. The summarizing logic is not in this skill but in the tubeless package (on PyPI); this skill is a thin wrapper that calls its CLI and relays the result to the user. A video with no captions, an LLM key/credit problem, and the like come back from the CLI as a one-line error -- relay that message as-is rather than throwing a stack trace at the user. ## Prerequisite This plugin calls the `tubeless` CLI, so it must be installed first: ``` pipx install tubeless # or: pip install tubeless ``` That puts the `tubeless` command on PATH. **tubeless finds the key itself** -- it reads `OPENAI_API_KEY` / `CLAUDE_API_KEY` / `GEMINI_API_KEY` from `~/.config/tubeless/credentials.json` (or the environment) per backend, so this skill never has to pull a key out and pass it. Never print a key value anywhere. ## Running Call `tubeless` from PATH with the `summarize` subcommand: ``` tubeless summarize "<URL>" [options] ``` (A bare `tubeless "<URL>"` also runs summarize as a shortcut, but the canonical form is `tubeless summarize`.) Options (`tubeless summarize --help` is the source of truth for exact defaults and models): - `--backend claude|openai|gemini|ollama` — the LLM vendor. Defaults to openai (or `backend` in `config.toml`). gemini has a free tier; ollama is local (no key). Each backend's default model is decided by the CLI -- do not hardcode a model here. - `--model <id>` — over