openclaw-skills

Central repository for all OpenClaw agent skills. Each folder contains a self-contained skill that OpenClaw can load and use.

Architecture

OpenClaw uses a classifier + sub-agent pattern:

                         User Request
                              │
                              ▼
                    ┌───────────────────┐
                    │   Main Agent      │
                    │   (Classifier)    │
                    │                   │
                    │ Reads SKILLS.json │
                    │ to decide which   │
                    │ sub-agent(s) to   │
                    │ activate          │
                    └──────┬────────────┘
                           │
              ┌────────────┼────────────┐
              ▼            ▼            ▼
     ┌──────────────┐ ┌──────────┐ ┌──────────┐
     │  crawlnode   │ │ passim…  │ │ (future) │
     │              │◄┤►         │ │          │
     │ Browse sites │ │ Upload & │ │ Calendar │
     │ Fill forms   │ │ share    │ │ Slack    │
     │ Screenshots  │ │ files    │ │ etc.     │
     └──────────────┘ └──────────┘ └──────────┘
           ▲                ▲
           └────────────────┘
           Skills collaborate
           with each other

1. Main Agent (Classifier) — receives every user request, reads SKILLS.json to understand what skills are available, then routes the task to the right sub-agent.
2. Sub-Agents (Skills) — each skill is a specialist. It handles one type of task really well.
3. Collaboration — skills can call on each other mid-task. For example, CrawlNode takes a screenshot and hands it to PassImageIn to upload and get a public URL.

Skill Registry

SKILLS.json at the root of this repo is the central index. It lists every skill, what it can do, what environment variables it needs, and which other skills it can collaborate with. The classifier reads this file to make routing decisions.

Skills

Skill	Folder	What it does
CrawlNode	crawlnode/	Remote browser automation — browse websites, interact with pages, capture screenshots
PassImageIn	passimagein/	Upload files to s.passimage.in and return public URLs for sharing

Folder Structure

openclaw-skills/
  SKILLS.json                         ← skill registry (classifier reads this)
  README.md                           ← this file
  crawlnode/
    SKILL.md                          ← skill definition (loaded by OpenClaw)
    docs/CRAWLNODE-API-DOCUMENTATION.md
    scripts/test-crawlnode.sh
  passimagein/
    SKILL.md                          ← skill definition (loaded by OpenClaw)
  (future-skill)/
    SKILL.md

Each skill folder must contain at minimum a SKILL.md file with:

• YAML front-matter (name, description, and optionally version, metadata)
• Instructions for how the agent should use the skill

Adding a New Skill

1. Create a new folder: my-new-skill/
2. Add a SKILL.md with YAML front-matter and usage instructions
3. Register it in SKILLS.json — add an entry with name, path, description, capabilities, required env vars, and which skills it can collaborate with
4. Push to GitHub
5. Run the refresh script on Claw1 to pull the update

Deployment

This repo is the single source of truth for all OpenClaw skills. On the Claw1 server:

~/scripts/refresh-openclaw.sh

This script pulls all skills from this repo into /home/autoscale/.openclaw/skills/ and restarts the gateway.

Related Repos

Repo	Purpose
asa-cf-worker.autoscale.team	ASA — Cloudflare Worker that connects Slack to OpenClaw