Back to Lobehub

lh task - Complete Command Reference

packages/builtin-skills/src/task/references/commands.md

2.1.564.0 KB
Original Source

lh task - Complete Command Reference

Core Commands

  • lh task list [--status <status>] [--root] [--parent <id>] [--agent <id>] [-L <limit>] [--tree] - List tasks
    • --status: pending, running, paused, completed, failed, canceled
    • --root: Only root tasks (no parent)
    • --tree: Display as tree structure
  • lh task view <id> - View task details (instruction, workspace, activities)
  • lh task create -i <instruction> [-n <name>] [--agent <id>] [--parent <id>] [--priority <0-4>] - Create task
    • Priority: 0=none, 1=urgent, 2=high, 3=normal, 4=low
  • lh task edit <id> [-n <name>] [-i <instruction>] [--status <status>] [--priority <0-4>] [--agent <id>] - Update task
  • lh task delete <id> [--yes] - Delete task
  • lh task clear [--yes] - Delete all tasks
  • lh task tree <id> - Show subtask tree with dependencies

Lifecycle Commands

  • lh task start <id> [--no-run] [-p <prompt>] [-f] [-v] - Start task (pending → running)
    • --no-run: Only update status, skip agent execution
    • -f, --follow: Follow agent output in real-time
    • -v, --verbose: Show detailed tool call info
  • lh task run <id> [-p <prompt>] [-c <topicId>] [-f] [--topics <n>] [--delay <s>] - Run/re-run agent execution
    • -c, --continue: Continue on existing topic
    • --topics <n>: Run N topics in sequence
  • lh task pause <id> - Pause running task
  • lh task resume <id> - Resume paused task
  • lh task complete <id> - Mark as completed
  • lh task cancel <id> - Cancel task
  • lh task comment <id> -m <message> - Add comment
  • lh task sort <parentId> <id1> <id2> ... - Reorder subtasks
  • lh task heartbeat <id> - Send manual heartbeat
  • lh task watchdog - Detect and fail stuck tasks

Checkpoint Commands

  • lh task checkpoint view <id> - View checkpoint config
  • lh task checkpoint set <id> [--on-agent-request <bool>] [--topic-before <bool>] [--topic-after <bool>] [--before <ids>] [--after <ids>] - Configure checkpoints
    • --on-agent-request: Allow agent to request review
    • --topic-before/after: Pause before/after each topic
    • --before/after <ids>: Pause before/after specific subtask identifiers

Review Commands (LLM-as-Judge)

  • lh task review view <id> - View review config
  • lh task review set <id> [--model <model>] [--provider <provider>] [--max-iterations <n>] [--no-auto-retry] [--recursive] - Configure review
  • lh task review criteria list <id> - List review rubrics
  • lh task review criteria add <id> -n <name> [--type <type>] [-t <threshold>] [-d <description>] [--value <value>] [--pattern <pattern>] [-w <weight>] [--recursive] - Add rubric
    • Types: llm-rubric, contains, equals, starts-with, ends-with, regex
    • Threshold: 0-100
  • lh task review criteria rm <id> -n <name> [--recursive] - Remove rubric
  • lh task review run <id> --content <text> - Manually run review

Dependency Commands

  • lh task dep add <taskId> <dependsOnId> [--type <blocks|relates>] - Add dependency
  • lh task dep rm <taskId> <dependsOnId> - Remove dependency
  • lh task dep list <taskId> - List dependencies

Topic Commands

  • lh task topic list <id> - List topics for task
  • lh task topic view <id> <topicId> - View topic messages (topicId can be seq number like "1")
  • lh task topic cancel <topicId> - Cancel running topic and pause task
  • lh task topic delete <topicId> [--yes] - Delete topic and messages

Document Commands (Workspace)

  • lh task doc create <id> -t <title> [-b <content>] [--parent <docId>] [--folder] - Create and pin document
  • lh task doc pin <id> <documentId> - Pin existing document
  • lh task doc unpin <id> <documentId> - Unpin document
  • lh task doc mv <id> <documentId> <folder> - Move document into folder (auto-creates folder)

Tips

  • All commands support --json [fields] for structured output
  • Task identifiers use format like TASK-1, TASK-2, etc.
  • Use lh task tree to visualize full task hierarchy before planning work
  • Use lh task comment to log progress — comments appear in task activities
  • Documents in workspace are accessible to the agent during execution