jammed-claw

name	jammed-claw
description	Create and run visual workflows using jam-nodes. Use when the user wants to build multi-step automations, data pipelines, or scheduled tasks as structured workflows instead of ad-hoc tool calls.

jammed-claw — Visual Workflows for OpenClaw

Build and run jam-nodes workflows that execute deterministically using OpenClaw tools. AI builds the workflow once → engine runs it forever with zero token cost.

Setup

1. Copy this skill/ folder into your OpenClaw workspace: ~/.openclaw/workspace/skills/jammed-claw/
2. Create a workflows/ directory in your workspace: ~/.openclaw/workspace/workflows/
3. Set environment variables for bridge features:
   • BRAVE_API_KEY — for web search node
   • OPENAI_API_KEY — for LLM node

When to Use

• User wants a repeatable multi-step automation (not a one-off task)
• Task involves chaining tools (search → summarize → send)
• User says "workflow", "pipeline", "automate", or wants something to run on a schedule
• A task would burn tokens on every run — make it deterministic instead

Workflow Format

Workflows are JSON files stored in workspace/workflows/. Structure:

{
  "id": "my-workflow",
  "workflow": {
    "name": "Workflow Name",
    "description": "What it does",
    "nodes": [
      { "id": "unique_id", "type": "node-type", "config": { "param": "value" } }
    ],
    "edges": [
      { "source": "node_1", "target": "node_2", "sourceHandle": "outputPort", "targetHandle": "inputPort" }
    ]
  },
  "createdAt": "ISO-timestamp",
  "updatedAt": "ISO-timestamp",
  "runCount": 0
}

Available Nodes

Node Type	Name	Inputs	Outputs
openclaw-web-search	Web Search	query (string), count (number, default 5)	results (array), resultText (string)
openclaw-web-fetch	Web Fetch	url (string), mode ("markdown"/"text")	content (string)
openclaw-exec	Shell Command	command (string), timeout (number, default 30)	stdout, stderr (string), exitCode (number)
openclaw-message	Send Message	target (string), message (string), channel? (string)	sent (boolean)
openclaw-llm	LLM	prompt (string), systemPrompt?, model?, temperature?, maxTokens?	response (string)
openclaw-memory-read	Read Memory	path (string)	content (string)
openclaw-memory-write	Write Memory	path (string), content (string)	written (boolean)
openclaw-conditional	Conditional	value (any), condition (string)	result (boolean), value (passthrough)
openclaw-template	Text Template	template (string), data (object)	text (string)
openclaw-merge	Merge	a, b, c? (any)	merged (object)
openclaw-json-extract	JSON Extract	data (any), path (string)	value (any)

Conditions (for conditional node)

Built-in (no LLM needed): not empty, empty, contains X, greater than N, less than N, equals X Complex/natural language conditions fall back to LLM evaluation.

Running Workflows

# List all saved workflows
node workspace/skills/jammed-claw/scripts/run-workflow.mjs

# Run a specific workflow
node workspace/skills/jammed-claw/scripts/run-workflow.mjs <workflow-id>

# Run with initial variables
node workspace/skills/jammed-claw/scripts/run-workflow.mjs <workflow-id> --var key=value

Creating a Workflow

1. Design the workflow JSON with nodes and edges
2. Save to workspace/workflows/<id>.json in the StoredWorkflow format above
3. Run it: node workspace/skills/jammed-claw/scripts/run-workflow.mjs <id>

Data Flow

• Edges pipe data between nodes: sourceHandle → targetHandle
• Use {{variable}} in config strings for interpolation from upstream node outputs
• Each node's output is stored as a variable named after the node's ID
• Output object keys are also stored as top-level variables

Conditional Branching

When a conditional node evaluates to false, all downstream nodes connected via edges are skipped.

Scheduling with Cron

Use OpenClaw cron to schedule workflow runs:

cron add → agentTurn → "Run the jammed-claw workflow: <id>"

Or as a system event that the agent picks up on heartbeat.

Example Workflow

"Search for AI news and summarize it":

{
  "id": "ai-news-digest",
  "workflow": {
    "name": "AI News Digest",
    "description": "Searches for AI news and summarizes the results",
    "nodes": [
      {
        "id": "search",
        "type": "openclaw-web-search",
        "config": { "query": "AI news today", "count": 5 }
      },
      {
        "id": "summarize",
        "type": "openclaw-llm",
        "config": {
          "prompt": "Summarize these search results into a brief news digest:\n\n{{resultText}}",
          "systemPrompt": "You are a concise tech news summarizer. Output 3-5 bullet points.",
          "temperature": 0.3
        }
      }
    ],
    "edges": [
      { "source": "search", "target": "summarize", "sourceHandle": "resultText", "targetHandle": "prompt" }
    ]
  },
  "createdAt": "2026-02-19T00:00:00Z",
  "updatedAt": "2026-02-19T00:00:00Z",
  "runCount": 0
}