stella scripts
List and run the project's package-manager scripts through Stella's canonical-verb index — the same frames the agent sees.
stella scripts lists and runs the project's package-manager scripts. Detection is deterministic static parsing of your manifests (cargo, npm/pnpm/yarn, uv, go, make, just, …), mapped onto canonical verbs — install, build, check, start, test, lint, format. It mirrors the list_scripts and run_script tools one-for-one, so a human at the CLI and the model inside a turn see the same script index. Offline: manifest parsing plus a local subprocess, no API key.
Synopsis
stella scripts list [--json] [--dir <path>]
stella scripts run <script> [--dir <path>] [--timeout-secs <n>] [-- <args...>]What it does
Instead of guessing that a repo uses pnpm test versus cargo test versus make check, Stella reads the manifests and presents one normalized index. The agent draws its verbs from exactly this index, which is why a run and a review agree on what "run the tests" means in your repo.
stella scripts list
List detected scripts and their canonical verb bindings.
| Flag | Meaning |
|---|---|
--json | Emit the index as JSON (schema_version 1) for tooling. |
--dir <path> | Narrow to one workspace package directory. |
stella scripts list
stella scripts list --json --dir apps/webstella scripts run
Run a script by canonical verb or by qualified id (e.g. test, pnpm:build).
| Flag / arg | Default | Meaning |
|---|---|---|
<script> | — | install | build | check | start | test | lint | format, or a qualified id like pnpm:build. |
--dir <path> | — | Package dir when the id exists in several packages. |
--timeout-secs <n> | 600 | Abort the subprocess after this many seconds. |
-- <args...> | — | Extra args appended runner-natively (after -- for npm-family runners). |
# Canonical verb
stella scripts run test
# Qualified id in a specific package, with extra args
stella scripts run pnpm:build --dir apps/web -- --no-lintThe script index is what makes verbs like test and build portable across repos. It also powers the agent's own run_script tool, so the CLI and the model never disagree about how your project builds.