| name | manus |
| description | Use Manus for async tasks that exceed local tools: broad research, deliverables such as PDF/PPT/CSV, connector-based work, or long multi-step jobs. Create a task, poll or receive webhooks, then fetch results with the bundled script. |
Manus
Prefer Manus only when local tools or sub-agents are not enough.
Docs: https://open.manus.im/docs
Rules
- Give Manus a complete prompt with scope, output format, and constraints
- Do not include secrets or unnecessary personal data
- Prefer
manus-1.6; use -lite for cheaper exploratory work
- Continue multi-turn tasks with
--task-id
Script
SCRIPT="<SKILL_DIR>/scripts/manus_client.mjs"
Create:
node "$SCRIPT" create \
--prompt "Your prompt" \
--mode agent \
--profile manus-1.6
Optional:
--attachment /path/to/file repeatable
--connector <uuid> repeatable
--task-id <id>
--label <text>
--locale zh-CN
--interactive
Status:
node "$SCRIPT" status --task-id <task_id>
Result:
node "$SCRIPT" result --task-id <task_id>
List:
node "$SCRIPT" list --limit 10 --status completed
Delete:
node "$SCRIPT" delete --task-id <task_id>
If Manus asks a follow-up question:
node "$SCRIPT" create \
--task-id <original_task_id> \
--prompt "User reply"
References:
- API:
<SKILL_DIR>/references/api.md
- Setup:
<SKILL_DIR>/references/setup.md
- Webhook helper (Node):
<SKILL_DIR>/scripts/webhook-transform.mjs
- Webhook helper (Python):
<SKILL_DIR>/scripts/webhook_transform.py
- Python fallback:
<SKILL_DIR>/scripts/manus_client.py