agentleFS
verbodh-pte-ltdverbodh-pte-ltd/motion-skills/AGENTS.md

motion-skills

AGENTS.md2 starsChanged 32 days agoMITView on GitHub

Instructions for any coding agent (Codex, Claude Code, Copilot, Gemini and others) and for the people who run them. This repository holds four skills that turn product work into short films and training material. All four live in skills/<name>/. Each folder has a SKILL.md with the full steps. Read that file before you start the task it covers. motion-web-video reuses the score synth of motion-video. Keep both folders side by side. - Claude Code: /plugin install motion-skills --marketplace verbodh-pte-ltd/motion-skills, or…

  • Reads credentials
  • Installs packages
  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository4 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

5 rules. About 1 is common, 2 are distinctive, and 2 are very specific to this project.

  • ≈1

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    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:

    All four live in skills/<name>/.

    line 9
  2. Distinctive:

    Read that file before you start the task it covers.

    line 10
  3. Very specific:

    Before the first film, run python scripts/check.py <skill>.

    line 32
  4. Common:

    Nothing private on screen or in a file: No keys, tokens, passwords, customer names, emails or security findings in a screenshot, a film or …

    line 71
  5. Distinctive:

    Ask, then build: Ask the user the two questions each film skill lists (style and colours) in one go, then build.

    line 78

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “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
AGENTS.md

AGENTS.md — motion-skills

Instructions for any coding agent (Codex, Claude Code, Copilot, Gemini and others) and for the people who run them. This repository holds four skills that turn product work into short films and training material.

The skills

All four live in skills/<name>/. Each folder has a SKILL.md with the full steps. Read that file before you start the task it covers.

SkillUse it when the user asks forIt makes
motion-videoa 10–30 s explainer, a product teaser, a motion clip for a posta silent MP4 from Manim scenes and a score written as notes
motion-web-videoan animated explainer with web-style motion, editable words, an optional voicean MP4 from a web page animated with Motion, plus a live editor
training-guidea write-up of a setup or a feature checked in a browsera guide in docs/training/<App>/<NN-Topic>/ with one screenshot per step
training-videoa video from a training guidea narrated MP4 with captions timed to the voice

motion-web-video reuses the score synth of motion-video. Keep both folders side by side.

Use them with your agent

  • Claude Code: /plugin install motion-skills --marketplace verbodh-pte-ltd/motion-skills, or python scripts/install.py to copy the folders into ~/.claude/skills/.
  • Codex: python scripts/install.py --codex --project <your project> (copies into .agents/skills/).
  • Other agents: give the agent the skill's SKILL.md as its instructions for the task.

In every SKILL.md, ${CLAUDE_SKILL_DIR} means that skill's folder. Claude Code fills it in; in other agents, read it as the folder that holds the SKILL.md.

Before the first film, run python scripts/check.py <skill>. It names each missing tool and the command that installs it on this OS. Tell the user what is missing; do not install tools without asking.

Tools to install

SkillNeeds
motion-videoPython 3, Manim Community (uv tool install manim), ffmpeg and ffprobe
motion-web-videoNode 22 or later, Google Chrome, ffmpeg and ffprobe, Python 3. No npm packages
training-guidea browser tool that can take screenshots (for example Claude in Chrome)
training-videouv (it brings Pillow and an ffmpeg binary)

ffmpeg: brew install ffmpeg on macOS, winget install --id Gyan.FFmpeg -e on Windows (open a new shell after it).

Settings

Drafts need no setting: they use the free system voice and write files locally.

VariableUsed byWhat for
ELEVENLABS_API_KEYmotion-web-video, training-videothe paid voice, only after the user approves a free draft
TRAINING_VIDEO_BUCKETtraining-videooptional S3 bucket for the voice cache and published videos, e.g. example-training-media
TRAINING_VIDEO_BASE_URLtraining-videooptional site that serves that bucket, e.g. https://www.example.com
TRAINING_VIDEO_PREFIXtraining-videooptional folder in the bucket, default media/training
AWS_PROFILE, AWS_REGIONtraining-videoread by the AWS CLI, only when a bucket is set
CHROME_PATHmotion-web-videothe Chrome executable, when it is not found

Rules for every agent

  1. Draft first, pay last. Make every film and video with the free system voice first. Use ElevenLabs only after the user approves the draft, and tell them the character count before the first paid run.
  2. Look before you report. Take a frame from the middle of each scene and view it. Check that the MP4 has sound: ffmpeg -i <file>.mp4 -af volumedetect -f null - must show Audio: aac and a mean volume above about −40 dB.
  3. Every claim on screen is true today. A feature that is only planned or only on a test site is not shown as live.
  4. Nothing private on screen or in a file. No keys, tokens, passwords, customer names, emails or security findings in a screenshot, a film or a commit. Read keys from the environment only; never print them or write them to a file.
  5. Use made-up names in shared examples: the product Acme Board at board.example.com, the bucket example-training-media.
  6. Never embed a narrated video as muted autoplay. Browsers autoplay only muted video, so the voice is never heard. Embed it with controls and a poster.
  7. Ask, then build. Ask the user the two questions each film skill lists (style and colours) in one go, then build. Do not ask again for what the request already says.

Check that the skills work

python skills/motion-video/scripts/test_motion.py       # score maths, cue timing, levels
python skills/motion-web-video/scripts/test_render.py   # renders the example film (Chrome, ffmpeg)
python skills/motion-web-video/scripts/test_voice.py    # free voice, clip reuse, --plan (ffmpeg)

The voice test never calls ElevenLabs.

Layout

motion-skills/
├── AGENTS.md                 this file
├── README.md                 overview and install
├── docs/                     INSTALL, QUICKSTART, TROUBLESHOOTING
├── scripts/                  check.py (tools per skill), install.py (copy the skills)
├── .claude-plugin/           plugin and marketplace files
├── LICENSE                   MIT
└── skills/
    ├── motion-video/         SKILL.md, scripts/motion.py, test
    ├── motion-web-video/     SKILL.md, example/, templates/, runtime/ (Motion, MIT), scripts/
    ├── training-guide/       SKILL.md, scripts/session_screenshots.py
    └── training-video/       SKILL.md, scripts/guide_video.py

Licences

MIT. skills/motion-web-video/runtime/motion.js is Motion (motion.dev), MIT, with its own licence file beside it. Do not copy Motion+ (paid) code into a film or a skill. Claude Motion is Anthropic's product; do not use its name or branding in published films.

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.