Multi-agent collaboration powered by OpenClaw cron jobs and the clawion CLI (wake-driven workflow).
Scanned 9/7/2026
Install to Claude Code
npx -y skills add modbender/skill-library-mcp --skill clawion --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Clawion?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/modbender-clawion)More formats (shields.io, HTML) on the badges page.
---
name: clawion
description: Multi-agent collaboration powered by OpenClaw cron jobs and the clawion CLI (wake-driven workflow).
---
# Clawion (runbook)
Clawion is a **file-based mission coordinator**. Agents interact with state through the **`clawion` CLI**.
Repo: https://github.com/natllian/clawion
## Mental model
Clawion is **wake-driven** — OpenClaw cron ticks are the engine.
1. Cron fires a periodic tick.
2. Agent runs **`clawion agent wake`** → receives the authoritative prompt for this turn.
3. Agent follows the **Turn Playbook** in that wake output.
4. Agent writes back via small actions (**message / working**).
5. Next wake reflects the updated workspace state.
Key properties:
- **Wake is the only read entrypoint.**
- **Unread handling is automatic:** wake shows "Unread Mentions" and auto-acks them.
## Core invariants
- **Workers complete tasks; managers maintain the board.**
- Worker: reports progress and asks questions via `message add`; logs process via `working add`.
- Manager: dispatches and maintains truth via `task create/assign/update`; also logs via `working add`. When the mission is complete, **disables all related cron jobs**.
- **Identity is explicit:** every action requires global `--agent <agentId>`.
---
## Quickstart (bootstrap a new mission)
### 0. Pre-flight
Verify the CLI is available:
```bash
clawion --help
```
If the command is not found, install it by following the setup instructions in the [Clawion README](https://github.com/natllian/clawion#readme).
### 1. Create the mission
```bash
clawion mission create --id <MISSION_ID> --name "..."
```
### 2. Register the manager (bootstrap rule)
The acting `--agent` must be the manager itself.
```bash
clawion agent add \
--mission <MISSION_ID> \
--id <MANAGER_ID> \
--name "Manager" \
--system-role manager \
--role-description "..." \
--agent <MANAGER_ID>
```
### 3. Register worker agents
```bash
clawion agent add \
--mission <MISSION_ID> \
--id <WORKER_ID> \
--name "Worker" \
--system-role worker \
--role-description "..." \
--agent <MANAGER_ID>
```
Repeat per worker.
### 4. Create and assign tasks (manager-only)
```bash
clawion task create \
--mission <MISSION_ID> \
--id <TASK_ID> \
--title "..." \
--description "..." \
--agent <MANAGER_ID>
clawion task assign \
--mission <MISSION_ID> \
--task <TASK_ID> \
--to <WORKER_ID> \
--agent <MANAGER_ID>
```
### 5. Write the roadmap (manager-only, one-shot)
```bash
clawion mission roadmap --id <MISSION_ID> --set "<markdown>" --agent <MANAGER_ID>
```
> The roadmap can only be written **once** via the CLI. Any subsequent edits must be made by a human in the Web UI.
### 6. Create cron jobs (disabled) and get user approval
Create **one isolated cron job per agent** (manager + each worker), all **disabled**. Then ask the user to review in the Clawion Web UI before enabling anything. See [Cron jobs](#cron-jobs-openclaw) for payload rules and naming.
**Web UI review checklist:**
- Mission ROADMAP (editable)
- Each agent's role description (editable)
- Dark Secret injection settings (ensure secrets don't leak into thread messages)
Only enable cron jobs after **explicit user approval**.
---
## Cron jobs (OpenClaw)
### Hard rules
| Rule | Detail |
| ------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Isolation** | Each tick runs in its **own isolated OpenClaw session** (never `main`). Context bleed makes the loop unreliable. |
| **Wake interval** | If the user didn't specify one, **ask and confirm** before creating jobs. |
| **Minimal payload** | Do **not** embed mission context, task lists, or SOP text. The authoritative prompt is assembled by `clawion agent wake` from workspace state at runtime. |
| **Disabled first** | Always create jobs disabled. Enable only after the user reviews in the Web UI (quickstart step 6). |
### Recommended cron message
**Worker:**
```text
Fetch your instructions by running:
clawion agent wake --mission <MISSION_ID> --agent <AGENT_ID>
Then follow the Turn Playbook in that output.
```
**Manager:**
```text
Fetch your instructions by running:
clawion agent wake --mission <MISSION_ID> --agent <AGENT_ID>
Then follow the Turn Playbook in that output.
If the mission is complete, disable all related cron jobs.
```
### Operational tips
- **Job naming:**
- `clawion:<MISSION_ID>:manager:<AGENT_ID>`
- `clawion:<MISSION_ID>:worker:<AGENT_ID>`
- **Stagger ticks** when multiple agents share the same interval to avoid bursty runs.
- Given interval = `N` minutes and `K` agents, choose offsets: `round(i * N / K)` for `i = 0..K-1`.
- Example: `N=10`, `K=3` → offsets `0m`, `3m`, `7m`.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!