Problem Remotion projects rely on a consistent set of best‑practice “skills” that AI agents can invoke. The community has a large skill library, but keeping it in sync with the CLI, generating the README and validating links is manual. The repo supplies a minimal, reproducible set of scripts and documentation to automate that process.

What This Does skills/* contains Markdown skill definitions (SKILL.md), agent YAML files (agents/openai.yaml) and optional assets. For example, skills/remotion-create/SKILL.md explains how to create a new composition, while skills/remotion-create/agents/openai.yaml declares the remotion-create prompt. Scripts in the scripts/ folder drive the maintenance workflow.

  • scripts/prepare-embedded-skills.ts copies skill artefacts into the embedded skills directory used by the Remotion CLI.
  • scripts/sync-agent-skills.ts scans the skills/ tree and updates internal metadata files.
  • scripts/sync-readme.ts regenerates the top‑level README from a template and the current skill list.
  • scripts/validate-links.ts checks that all URLs in the documentation resolve.

These scripts are written in TypeScript, so tsconfig.json drives compilation and the package.json lists typescript and ts-node as devDependencies.

How It Is Wired Execution starts at scripts/sync-readme.ts.

generateReadme() → write README.md

generateReadme() calls collectSkills() (in scripts/sync-agent-skills.ts) which walks skills/, reads each SKILL.md, and builds a Markdown table. prepareEmbeddedSkills() then copies the assets from each skill into packages/skills/skills/ (the path the Remotion CLI consumes). The only external effect is filesystem writes: README.md, packages/skills/skills/*, and a packages/skills/skills/skills.json index. No network calls or database interactions occur. Because each script imports only local modules, the graph is acyclic. The widest blast radius is prepare-embedded-skills.ts, which touches every skill asset; changing that file would force a full rebuild of the embedded skill bundle.

How To Use It

  1. Install ``bash npm ci # or pnpm install ` package.json declares the dependencies; npm or pnpm` is sufficient.
  2. Sync the skill set ``bash npx ts-node scripts/sync-agent-skills.ts npx ts-node scripts/prepare-embedded-skills.ts npx ts-node scripts/sync-readme.ts ` After these steps, the packages/skills/skills/` directory contains the embedded skills and the README reflects the current list.
  3. Add the skills to a Remotion project ``bash npx remotion skills add ` or, if Remotion is not installed, use the generic CLI: `bash npx skills add remotion-dev/skills ``

Real‑World Use A product team writes a prompt such as /remotion-create Build a promo for a record store. The agent resolves it to the remotion-create skill, which the CLI maps to the Markdown definition and optional agent YAML. The CI pipeline of the product repo can run scripts/validate-links.ts to guarantee that the skill documentation remains link‑clean before a release.

Code Health & Issues

  • Medium/SDLC – No test files (.test.*) → untested code paths.
  • Medium/SDLC – No CI/CD config (.github/, ci.yml) → no automated gate.
  • Low/Risk – package.json missing a lockfile → non‑reproducible builds.

All findings are based on static inspection of the file tree; the repository contains no lint or type‑check errors.

The Bottom Line Remotion-skills is a lightweight, well‑structured repository that bundles a large collection of AI‑friendly Remotion skills and tooling to keep them in sync. Its main value lies in the documentation and build scripts; the code itself is straightforward and side‑effect free. Teams that need a curated skill set for AI agents, and who are comfortable adding a few TypeScript scripts to their workflow, will benefit most.