voiceover — Claude Code skill: demo script first, then code
The voiceover skill writes the demo narration before any code exists, then carries the feature through to a pull request with frame-by-frame proof attached. Useful when the team should agree on what a viewer sees first.
LicenseRef-Proprietary ★ 7,523 updated 2026-10-29
npx skills add Devin-AXIS/iPolloWork --skill voiceover -a claude-codeWhat voiceover does
Flip the usual order: narration first, implementation after. The agent asks a few questions about the feature, then drafts the demo voice-over frame by frame — one numbered paragraph per screen, 4–8 of them for most features. You edit words instead of diffs until you would genuinely read it aloud. Nothing gets built before you sign off on the script.
On approval it creates a separate checkout with git worktree add from origin/dev, saves the script to evals/voiceovers/<flow-id>.md, and runs pnpm fraimz scaffold <flow-id> to generate evals/flows/<flow-id>.flow.mjs with a ctx.prove stub per paragraph. Renumbering or rewording paragraphs later needs a fresh approval, because a flow fails the check once its narration no longer matches the file.
The executor subagent writes the feature; the fraimz skill loop drives the demo against the running app, repairs what broke and re-runs until every frame passes. Shipping means gh pr create plus pnpm fraimz --flow <flow-id> --pr, which drops the per-frame evidence onto the pull request as a comment. You need pnpm, gh and fraimz installed, or phases three and four stall. The repository is proprietary-licensed, so check the terms before wiring it into a team repo.
What it helps with
- agree on a demo before anyone writes code
- rewrite narration until it sounds speakable
- scaffold a proof flow from the approved script
- drive the demo against the running app and fix what breaks
- attach per-frame evidence to the pull request
Example prompts for the agent
- Write the voice-over for the CSV export feature, five frames.
- Run
/voiceoverand redo frame two — it describes the code instead of what the viewer sees. - Scaffold the flow from the approved script and run it until every frame passes.
How to install voiceover
1Skills CLI (Vercel)
npx skills add Devin-AXIS/iPolloWork --skill voiceover -a claude-code2Manually: put the files in place
~/.claude/skills/voiceover/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/voiceover/
1Skills CLI (Vercel)
npx skills add Devin-AXIS/iPolloWork --skill voiceover -a codex2Manually: put the files in place
.agents/skills/voiceover/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add Devin-AXIS/iPolloWork --skill voiceover -a cursor2Manually: put the files in place
.cursor/skills/voiceover/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install Devin-AXIS/iPolloWork .opencode/skills/voiceover2Manually: put the files in place
.github/skills/voiceover/SKILL.md1Skills CLI (Vercel)
npx skills add Devin-AXIS/iPolloWork --skill voiceover -a gemini-cli2Platform CLI
gemini skills install https://github.com/Devin-AXIS/iPolloWork --path .opencode/skills/voiceoverManually: put the files in place
.devin/skills/voiceover/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about voiceover
How do I install voiceover in Claude Code?
Run npx skills add Devin-AXIS/iPolloWork --skill voiceover -a claude-code; the skill file lands at ~/.claude/skills/voiceover/SKILL.md.
What does voiceover do?
It turns a feature request into a spoken demo script and follows it through to the PR: isolated worktree, build, frame runs, proof posted as a comment.
What has to be in place to reach the PR?
pnpm, gh and the fraimz skill, which runs the check loop and publishes the frame evidence. Coding is delegated to the executor subagent.