sprintsai/agent

This skill has three subcommands: launch, stop, and settings.

Check provider settings first

Before launching, always check which providers are configured:

lightsprint agent settings --output json

If you need environment IDs (required for codex, optional for anthropic):

lightsprint agent settings --provider codex --output json

Launch a cloud agent

Launch one or more agents. Use multiple --task flags to launch in parallel:

lightsprint agent launch --task <taskId> [--task <taskId> ...] --provider <anthropic|cursor|codex> --output json

Examples:

Single task:

lightsprint agent launch --task LS-100 --provider anthropic --output json

Multiple tasks (launched concurrently):

lightsprint agent launch --task LS-100 --task LS-101 --task LS-102 --provider anthropic --output json

Optional flags:

• --model <model> — override the provider's default model
• --base-ref <branch> — base branch (defaults to repo's default branch)
• --environment-id <id> — environment for codex (required) or anthropic (optional)

Important:

• --provider is always required
• Codex typically requires --environment-id — use agent settings --provider codex to discover IDs
• Only one agent can run per task per provider at a time
• When launching multiple tasks, all launches run concurrently and results are returned as an array

Stop an active agent

lightsprint agent stop --task $ARGUMENTS --provider <anthropic|cursor|codex> --output json

This interrupts the currently running agent for the task. The agent record is preserved for audit.