Create, version, schedule, and run declarative JSON workflows with AI, agent, and deterministic nodes.
npx skills add https://github.com/clawnify/clawflow --skill clawflow-clawnify
Please help me install this Agent Skill. Skill: clawflow Source: https://github.com/clawnify/clawflow/tree/main/skills/clawflow Command: npx skills add https://github.com/clawnify/clawflow --skill clawflow-clawnify
flow_create, flow_edit, flow_publish, flow_run, and flow_trigger. - 12 node types: ai (structured LLM calls with schemas and attachments), agent (delegation to OpenClaw agents), exec, http, code, loop, branch, condition, parallel, memory, wait (approval gates and events), and sleep. - Versioning and scheduling: draft/publish versioning with immutable snapshots, plus cron-based triggers with per-trigger inputs and timezone support. - Use Case: Build a content pipeline that researches a topic with an AI node, drafts a post, pauses for human approval via a wait node, then publishes through an HTTP call with retries. ## Quick Start Ask the agent to create a flow that drafts a LinkedIn post from a topic using an AI node with a schema, then run it with flow_run.High-intent search queries and answers about installing and using this skill.
Define a JSON flow with a name and a nodes array, then use flow_create to save it and flow_run to execute it. Each node needs a unique name and a do type such as ai, exec, or http, and results are shared between nodes through output keys referenced in templates.
Add a wait node with for set to approval before any irreversible side effect. The flow pauses, shows the approver a prompt and optional preview data, and resumes via flow_resume with the instance ID and an approved flag.
An ai node makes a single LLM call with optional structured schema output, while an agent node delegates a task to a real OpenClaw agent with tool access like browser, exec, and memory. Use ai for extraction and generation, and agent for tasks needing tools.
Yes, flow_trigger creates cron-based triggers with standard 5-field cron expressions and optional IANA timezones. Triggers run the latest published version by default, support per-trigger inputs, and skip runs when the previous execution is still active.
Templates reference the node's output key, not its name, and the output field is required to store any result in state. If a node lacks output, its result is discarded, so add output to every node whose data downstream nodes reference, including loop and branch nodes.
Yes, flows use draft/publish versioning where flow_publish promotes the draft to an immutable numbered version. flow_run executes the latest published version by default, and you can pin specific versions or test drafts explicitly.
jenticAccess external APIs securely through the Jentic broker with centralized credentials.
PostHogManage PostHog subscriptions for scheduled insight and dashboard snapshots.
Mark393295827Design production agent harnesses with permissions, observability, and recovery paths.
attrix-technologiesBuild automated fleet monitoring workflows in n8n connecting Geotab data to Slack, Discord, and email.