monorepo-dev-workflow
Skill35 starsChanged 14 days agoMITView on GitHub
Use when building, testing, linting, formatting, typechecking, or running any app or package in the ffmpeg-video-composer monorepo, or when looking for the right pnpm/vp command.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- authoring-video-templatesSkill
- core-architecture-patternsSkill
- cross-platform-ffmpegSkill
- ondevice-ffmpeg-engineSkill
- pr-creationSkill
- pr-evidenceSkill
The file
Files
Monorepo Dev Workflow
Overview
pnpm workspaces (apps/*, packages/*, plus examples/llm-remotion-title), no turbo/nx. Tooling is vite-plus (vp) — it provides lint (oxlint), format, test (vitest), and staged checks. There is no eslint, no prettier, and no jest at the root (jest lives only inside apps/leclap-expo).
- pnpm 12.6.0, Node ≥ 24.11.0 (
engine-strict— wrong versions are rejected). Install:pnpm install.
Command map
| Task | Command |
| Lint | pnpm lint |
| Format / check | pnpm fmt · pnpm fmt:check |
| Format/lint/type checks | pnpm check |
| Package tests | pnpm test (recursive; Vitest packages + Expo Jest) |
| Build | pnpm build (recursive; tsdown libraries/CLI/MCP + Vite web app) |
| Typecheck a package | pnpm --filter <pkg> exec tsc --noEmit |
| Perf bench / profile | pnpm --filter ffmpeg-video-composer bench · FVC_PERF=1 pnpm compile <t.json> (docs/performance.md) |
| Run Expo app | pnpm app:expo · app:ios · app:android |
| Run web app | pnpm app:web |
| Build executables | pnpm build:exe:all |
| Dep graph / check | pnpm deps:graph (requires Graphviz dot) · pnpm deps:check |
<pkg> names: ffmpeg-video-composer, @leclap/cli, @leclap/mcp, @leclap/creative-kit, @leclap/expo, @leclap/web, leclap-json-effects-example.
Scope and prerequisites
pnpm test:ui,pnpm test:coverage, andpnpm test:ciuse the root Vitest config, which covers core, MCP, and repo-level tests. They omit CLI, creative-kit, web, and Expo package suites.pnpm testruns workspace packages' own test scripts; it does not collect repo-leveltests/.- CI runs root coverage, web/Expo/CLI tests separately, and
pnpm test:integrationfor core/MCP Cucumber real renders. Rust tests run in a separate host-engine job; web Playwright tests are a separatepnpm --filter @leclap/web test:e2ecommand. - Web Playwright tests reuse a dev server at
E2E_BASE_URL(defaulthttp://localhost:5174);E2E_CHROMIUM_PATHpoints at a preinstalled Chromium. The browser-agent (WebMCP) suite runs without WASM through the dev polyfill:pnpm --filter @leclap/web exec playwright test e2e/webmcp-builder.spec.ts;E2E_WASM=1adds the preview and frame cases. Seedocs/webmcp.md. pnpm checkchecks formatting/lint/types; tests and builds are separate.pnpm buildincludes the web app's media fetch/staging and prerender, but does not build Expo or the Rust engine. Filter the affected package for a narrower build.- Fresh-checkout lint needs generated imports: run
pnpm --filter ffmpeg-video-composer buildandnode scripts/copy-core-assets.tsfirst (the CI prepare step). Devpnpm compile/pnpm diagnosealso import the core's builtdist. - Render/probe tests require real media and FFmpeg with
drawtextplusffprobe. CI usesscripts/ci/fetch-test-media.shand verifies its allowlist; Git LFS pointer files alone are not valid test media. See.github/workflows/ci.ymlfor the current setup.
Before committing
pnpm fmt— format (semicolons, single quotes, width 120, 2-space, trailing comma es5).pnpm lint— oxlint must be clean.pnpm testfor affected code;tsc --noEmiton the package you changed.
Git hooks run vp staged checks automatically (fmt on most files, lint on *.{ts,tsx}), but run them yourself first to avoid hook failures.
Common mistakes
- Reaching for
eslint/prettier/jestconfigs at the root — they don't exist; usevp/ the table above. - Running raw
tscwithout--filter/-p— typecheck per package. - Ignoring
engine-strict— install fails on the wrong Node version; switch Node (e.g. viamise/nvm) rather than forcing the install.
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
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.

