agentleFS
tandryukhatandryukha/aidemo/.claude/skills/record-demo/SKILL.md

record-demo

Skill32 starsChanged 7 days agoMITView on GitHub

Produce a short (30-60s) narrated, captioned product demo video with the aidemo engine, via the aidemo MCP server. Use when the user says "record a demo of X", "make a demo video", "demo this feature", or points at a flow (e.g. a shopping flow inside a ChatGPT app) and wants a shareable MP4 with voiceover + captions. The skill authors a storyboard.json (script + per-scene voice/music plan + browser action-spec) and renders it with the engine. Also covers native iOS/Android apps and store media: "make an App Store preview", "Play promo video", "store screenshots" — record the iOS Simulator / Android emulator with Maestro flows and export App Store / Google Play videos and screenshot sets (guide topic `store`). Do NOT use for: live screen-sharing, editing an already-rendered MP4 you have no project for, or physical devices (simulators/emulators only).

  • Reads credentials

15 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 15 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Fallback — no MCP server registered: run aidemo guide and follow the same document using the CLI commands it maps out.

    line 39
  2. Distinctive:

    Put what you watched in your report.

    line 51
  3. Very specific:

    If unsure, run aidemo doctor (or the MCP doctor tool) — it prints the engine version and checks prereqs (ffmpeg, Chrome, OPENAI_API_KEY — …

    line 66

“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

record-demo

You are the demo director. Turn a feature + loose directions into a polished browser demo video using the aidemo engine.

This file is a thin adapter. The authoring knowledge lives in the engine — fetch it first, then follow it:

  1. Preferred — the aidemo MCP server (registered by aidemo repo-init, or bundled automatically if you installed the Claude Code plugin): call the get_authoring_guide tool and follow the guide (topic: "store" for native apps / App Store / Google Play media). Pipeline operations (probe, render, voice, captions, compose, gif) are jobs: they return a jobId immediately — poll job_status for progress, results, and failure artifacts. Discover selectors with inspect (unique selectors per visible element — no guessing) before writing targets; validate storyboard edits with validate_storyboard and lint_storyboard (pacing forecast: which scenes will freeze or be cut) before rendering; read output/report.json after. Pass absolute demo directories.
  2. Fallback — no MCP server registered: run aidemo guide and follow the same document using the CLI commands it maps out.
  3. Watch it as a viewer before you call it done. Play it at 1x, muted, as a stranger on a phone would see it, and answer the checklist in the guide's "Watch it as a viewer first". A valid report is not a watchable video. The usual misses are the ones a viewer notices at once:

    • too many scenes for the runtime (aim for 5–6 s per captioned scene, and drop scenes rather than speeding up);
    • results and captions gone before they can be read;
    • the key moment flashing past;
    • captions sitting on the app's own text.

    Put what you watched in your report. Never hand a cut over unwatched.

  4. After verifying the final video (per the guide's "Verify before declaring done"): if you hit any engine bug, surprise, or workaround this session, file it now — the MCP feedback tool (title + body; environment/log context is auto-attached) or aidemo feedback <dir> on the CLI (falls back to a local docs/feedback-*.md offline). If nothing came up, skip this step.

Invoking the engine (aidemo)

aidemo <cmd> means whichever of these applies:

  • In the aidemo repo itself (a full checkout): node bin/aidemo.mjs <cmd>.
  • In a product repo where this skill was installed via aidemo repo-init (the engine is not vendored here): npx -y github:tandryukha/aidemo#stable <cmd>.

If unsure, run aidemo doctor (or the MCP doctor tool) — it prints the engine version and checks prereqs (ffmpeg, Chrome, OPENAI_API_KEY — or OPENAI_BASE_URL pointing at a local OpenAI-compatible server, or AIDEMO_TTS_PROVIDER=local for in-process TTS + offline captions; in both cases no key is needed and doctor reports what's in effect).

Distribution & updates

This skill is a versioned copy of the source in the aidemo repo. In a product repo it arrived via aidemo repo-init, which also added a SessionStart hook that runs aidemo skill check and prints a notice when a newer version is out. (The authoring guide itself never goes stale — it is served by the engine at render time.)

  • Update the skill (when notified, or any time): aidemo skill update --dir . — rewrites this file and bumps .claude/skills/record-demo/installed.json. Installed as a Claude Code plugin instead? Update it with /plugin (the marketplace flow), not aidemo skill update.
  • Send feedback upstream: see step 4 above.
  • Bootstrap a new repo: npx -y github:tandryukha/aidemo#stable repo-init --dir .

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.