Mux Video

Mux docs discovery

Mux publishes its entire documentation as agent-ready markdown, auto-generated from the same source as the real docs site. For any Mux question, fetch the relevant page and answer from it — never answer Mux API questions from memory, because API shapes and guidance change.

Workflow

1. Route — try these in order and stop at the first hit. The order is cost-ordered: each step reads fewer tokens than the next.
   1. If the mux CLI is installed, run mux docs find "<topic>" --agent — it searches the docs index CLI-side and returns only the matching page URLs, so you never read an index into context. (Check mux docs --help; older versions lack it.)
   2. Construct the URL from a known pattern: the tables below (collections, Robots robots-<task>.md, API references) cover most topics.
   3. Fetch the smallest collection index for the topic area (a few dozen lines).
   4. Fetch https://www.mux.com/llms.txt — the master index of every docs page — only when the steps above fail.
2. Fetch ONE page — the page's markdown version: append .md to its URL (e.g. https://www.mux.com/docs/guides/start-live-streaming.md). Only add an API reference bundle when you need exact request/response shapes; do not fetch extra pages "for context."
3. Answer from the fetched content and cite the page URL.

Never fetch https://www.mux.com/llms-full.txt. It is the entire documentation in a single file and will flood the context window; every page in it is reachable individually through the steps above.

Collection indexes

Mux Robots (AI video workflows)

Robots — Mux's AI-powered video workflows — don't have their own collection index; their guides are part of the video collection and the master llms.txt index. Start here:

Task-specific guides follow the pattern https://www.mux.com/docs/guides/robots-<task>.md, where <task> is one of: summarize, moderate, generate-chapters, ask-questions, find-key-moments, find-scenes, find-best-thumbnails, translate-captions, translate-audio, generate-premium-captions, edit-captions, generate-engagement-insights.

API references and specs

For exact endpoint, parameter, and webhook shapes, prefer these over prose guides:

All of these resources are also discoverable from the docs site itself: every docs page links to "Docs for LLMs" (https://www.mux.com/docs/core/llms-txt), the canonical page describing Mux's llms.txt files and machine-readable bundles. If this skill's URL lists ever drift from reality, that page is the source of truth.

Self-healing

If a docs URL 404s, the docs have likely been reorganized. Do not give up after one 404: run mux docs find "<topic>" --agent (or fetch https://www.mux.com/llms.txt and search it) to get the page's current URL.

Staying in sync

• This skill is distributed from the muxinc/skills repository and ships embedded in the mux CLI. Updating the CLI (brew upgrade mux / npm update -g @mux/cli) updates the embedded skills, and mux skills update refreshes all local copies (including ~/.claude/skills) to match the installed version. For skills decoupled from CLI releases, install the mux@mux Claude Code plugin or pull directly from https://github.com/muxinc/skills.
• The CLI embeds only skill instruction files, never docs content — always fetch docs live using the workflow above. Check mux skills --help for what the installed version offers.

Guardrails

• Cite the docs page URL you answered from.
• If the network is unavailable, say you could not verify against current docs — never guess API request/response shapes from memory.