npm Script Naming
A script is named by who runs it: a task as <verb>:<app>:<thing>, and an automation context's script as <context>:<thing>.
Why
A script's name says what it runs and for which context.
Tasks (Run By Humans Or Shared)
Format: <verb>:<app>:<thing>. App always comes second, never <verb>:<thing>:<app>. Omit <app> in single-app projects. Omit <thing> when not needed.
Verbs
A verb names the kind of run, such as these:
start:- long-running processes (dev server, report viewer).test:,lint:,build:,e2e:,install:- one-shot runs.
Examples
start:admin- run the admin dev server.e2e:admin:report- run the full e2e suite with HTML report and videos.
Contexts (Run By Automation)
Format: <context>:<thing>. Each automation context gets its own entry scripts, distinct from the human-run tasks above. The hooks: and ci: prefixes are examples of the general pattern: one prefix per context.
An entry script calls a task, such as "hooks:pre-commit": "npm run lint", so a context and a human run one command that cannot drift. When a context needs a command of its own, that command is a task as well, such as lint:ci, and the entry script calls it.
hooks:pre-commit,hooks:pre-push- called by git hooks.ci:lint,ci:test- called by CI workflow steps.
Wrappers
For tools that ship as installable binaries (Playwright, MSW, etc.), add a thin wrapper script in package.json: "install:playwright": "playwright install --with-deps chromium". Call it via npm run install:playwright rather than invoking the binary directly.
Don't expose internal env vars or build flags (DEMO=1, CI=true) in script names. Keep them inside the script.