If rehearsal, transcript checks and uploading metadata are stealing time from the creative work that actually moves your channel forward, you need a workflow that removes those repetitive tasks. This set of seven prebuilt “Claude skills” turns research, post-shoot quality control, a first AI rough cut, packaging and upload prep into repeatable commands you run inside Anthropic’s Claude—so you spend more time shooting and refining and less on grunt work.
What these skills are and how to install them
A “Claude skill” is a bundled instruction set (a folder or zip) that tells Claude to perform one specific job when invoked. These skills are designed for Claude Fable 5.2 and can be installed either in Claude Code or the Claude app’s custom skills area.
Quick install checklist:
- For Claude Code: unzip the download and place each skill folder in your skills directory (for example ~/.claude/skills/).
- For the Claude app: open Customize > Skills and upload each skill zip you want to use.
- Choose Claude Fable 5.2 as the model when running the skills.
- Some steps call external services: vidIQ for search-volume data and Kapwing for AI editing and dubbing. Verify account access and any credit limits before you rely on the workflow.
Each skill writes its outputs into one folder per video so the next skill can pick up the files it needs. A Claude plan that supports custom skills is required to use them.
The seven skills and where to run them in your process
Below are the skills in the order they’re most useful for producing a typical tutorial or how‑to video. The descriptions focus on what they automate, when to run them and the manual checks you should still perform.
- /outliers — idea discovery: Runs keyword lookups (via vidIQ) and searches YouTube for phrasing variants, then ranks videos that have many more views than their channel size. That surfaces ideas, title formats and thumbnails that attract external traffic without a large subscriber base. Run this before you commit to a topic; it produces a keyword table and a shortlist of high-performing ideas to inspire angles and hooks.
- /youtube-script-researcher — research brief: Pulls transcripts and comments from top videos on your topic and synthesizes what each video covers, the hooks used in the first 30 seconds, gaps in coverage and common viewer questions or complaints. Use this after you pick an idea so your script answers what viewers are actually asking for.
- /footage-check — immediate post-shoot QC: Matches separately recorded audio to camera footage by waveform, transcribes takes (Kapwing), and flags missed lines or bad takes with timestamps so you can re-record before you leave the set. This keeps pickups cheap and prevents surprises in edit.
- /kai-rough-cut — first edit pass: Uploads raw footage to Kapwing and asks Kapwing’s Kai agent to produce a rough cut that preserves the best take of each scripted line while removing dead air and obvious flubs. It creates a Kapwing project with editable clips—treat this as a time-saving scaffold, not a finished edit: review cut points and pacing manually.
- /packaging — title and thumbnail ideas: Uses outliers output and the research brief to generate title options and thumbnail concepts (including image-generation prompts for Kapwing). It ranks thumbnail concepts against your channel’s typical performance; use these as starting points and refine them to fit your brand voice.
- /upload-pack — upload-ready metadata: Transcribes the finished video, drafts title options, description copy, timestamped chapters, suggested tags, an SRT subtitle file and a pinned comment. The output is formatted for pasting into YouTube Studio—verify chapters, timestamps and tone before publishing.
- /community-post — social hooks: Produces a YouTube community post carousel and captions aligned with the packaging and upload copy so your in‑platform promotion matches the published video.
Bonus: /make-it-global — a dubbing workflow that generates multi‑language audio tracks in your voice, prepared for YouTube’s multi‑language audio uploads. This relies on Kapwing’s dubbing features and credits.
Practical considerations and workflow tips
- These skills automate repetitive steps, not editorial judgment. Expect to spend time reviewing AI outputs—especially cuts, chapter labels and thumbnails—before publishing.
- AI transcripts intentionally preserve disfluencies and restarts so you can spot where pickups are needed; they won’t look like cleaned prose until you edit them.
- Validate account access, data-sharing settings and credit limits for vidIQ and Kapwing before scaling this workflow. Service changes or API limits will break parts of the chain.
- Check chapters and timestamps carefully: they set viewer expectations and affect discoverability. Errors here harm the watch experience even if the video itself is solid.
- Use outliers as directional signals, not prescriptions. High view-to-subscriber ratios reveal formats that attract external traffic, but you must adapt those formats to your niche and brand voice.
To test the approach: run /outliers on a topic you already know, then follow the chain through /youtube-script-researcher, /footage-check and /kai-rough-cut on a single video. Monitor time saved on research and initial edits, and identify where manual review still produces the biggest gains.
What to watch next: updates to Claude’s custom skills support, changes to Kapwing’s credit model or vidIQ’s data access, and YouTube’s policies or tooling for multi‑language audio. If you’re planning to scale, map account permissions and data flows before rolling these skills into a multi‑person production process—those operational details determine whether automation actually speeds you up or just shifts the bottleneck.