name: long-run-runtime description: Operate AutoForge as a durable long-running agent runtime with queueing, watch mode, async messages, checkpoint replies, and resumable execution.
Long-Run Runtime
Use this skill when the user needs a workflow that should continue beyond the current Claude session.
AutoForge supports:
- queue-based background execution
- periodic progress updates
- async user interference through messages
- checkpoint/bridge responses
- paused runs with later resume
Canonical runtime flow:
- Start daemon if not already running:
autoforgeai daemon start
- Queue work:
autoforgeai queue "<objective>" --wait --tail
- Inspect and follow:
autoforgeai projectsautoforgeai watch <project_id> --tail
- Interfere asynchronously:
autoforgeai msg <project_id> "<updated guidance>"
- Resume if paused:
autoforgeai unpause <project_id>
Use this mode whenever the user cares about:
- not losing state
- seeing incremental output
- injecting late feedback
- having one runtime pattern for software, research, and verification tasks