task-runner

name	task-runner
description	Run monorepo tasks with dependency ordering using monorepo-task-runner. Covers mtr run, mtr status, tasks.yaml format, and the web dashboard.

task-runner skill

Use this skill when the user wants to:

• Run one or more tasks with dependency resolution
• View current task status
• Understand tasks.yaml configuration
• Use the web dashboard to monitor a run

Prerequisites

# Server must be running (starts automatically with mtr run)
node dist/mtr.js serve
# or just run tasks directly
node dist/mtr.js run --all

tasks.yaml format

tasks:
  <name>:
    command: <shell command>      # required
    depends_on: [task1, task2]    # optional - other task names
    inputs:                       # optional - glob patterns for cache key
      - "packages/*/src/**/*.ts"
    outputs:                      # optional - files to cache/restore
      - "packages/*/dist/**"
    env:                          # optional - extra env vars
      NODE_ENV: production
    cwd: packages/server          # optional - working directory
    timeout: 300000               # optional - ms (default 300s)

Running tasks

# Run all tasks
mtr run --all

# Run specific tasks (includes their dependencies automatically)
mtr run build

# Run multiple tasks
mtr run lint typecheck

# Skip cache for a full re-run
mtr run --all --force

# Custom config file
mtr run --config ./ci/tasks.yaml --all

Task status

mtr status

Output columns: TASK, STATUS, LAST RUN, DURATION, CACHE

Status values:

• success - completed successfully
• failed - exited non-zero
• cached - skipped, cache hit used
• skipped - not run because a dependency failed
• running - currently executing
• pending - waiting for dependencies

Dependency resolution

Tasks are executed in topological order. Tasks at the same stage (no mutual dependency) run concurrently up to MTR_CONCURRENCY (default 4).

Example for tasks.yaml with test -> build -> [lint, typecheck]:

• Stage 1: lint and typecheck run concurrently
• Stage 2: build runs after both pass
• Stage 3: test runs after build passes

If any task fails: no new tasks start, running tasks finish, dependents marked skipped.

Web dashboard

Open http://127.0.0.1:3300 to see:

• Pipeline view: stage grid with task cards, live status updates via SSE
• Run history: all past runs with status and duration
• Task detail: per-task run history, cache hit rate
• Cache view: list entries, delete individual or all

The dashboard auto-refreshes during active runs. No manual reload needed.

Environment variables

MTR_PORT=3300           Web UI port
MTR_HOST=127.0.0.1      Bind address
MTR_CACHE_DIR=./.mtr-cache   Cache location
MTR_CONCURRENCY=4       Parallel task limit
MTR_WEB_ENABLED=1       0 to disable web UI

Exit codes

• 0: all tasks passed
• 1: one or more tasks failed
• 2: config error (bad tasks.yaml, circular dependency)