🎵 SKIO Contest Infrastructure — Starter Kit

This is your AI-powered contest production engine. It gives your OpenClaw agent everything it needs to plan, create, and launch SKIO remix contests end-to-end — from brief to postmortem. Just fill in your knowledge files, drop in a contest brief, and let the agent do the heavy lifting.

🚀 Setup (5 minutes)

Prerequisites

• OpenClaw installed and running
• Connected to Discord (your SKIO server)
• You can chat with your agent

Step 1: Copy to your workspace

cp -r skio-starter-kit/* ~/.openclaw/workspace/

This merges the kit into your workspace. Your agent will automatically read AGENTS.md and SOUL.md on every session.

Step 2: Fill in your knowledge files

The knowledge/ folder is your agent's brain. The more you fill in, the better it performs.

Start with these (most impactful first):

1. knowledge/brand/voice-guide.md — How SKIO sounds
2. knowledge/strategy/skio-philosophy.md — What makes SKIO tick
3. knowledge/strategy/producer-psychology.md — What motivates your producers

💡 Tip: Each file has prompting questions to help you fill it in. You can even paste answers into Discord and ask your agent to organize them into the file.

Step 3: Say hi

Tell your agent: "Read your AGENTS.md and let me know you're ready."

That's it. You're set up.

🔄 How It Works

┌─────────────┐     ┌──────────────┐     ┌──────────────┐
│  YOU drop a  │────▶│  Agent runs   │────▶│  You review  │
│ contest brief│     │  the pipeline │     │  & approve   │
└─────────────┘     └──────────────┘     └──────────────┘
                           │
                    ┌──────┴──────┐
                    ▼             ▼
              ┌──────────┐ ┌──────────┐
              │ strategy │ │ creative │
              │   .md    │ │   .md    │
              └────┬─────┘ └────┬─────┘
                   ▼            ▼
              ┌──────────┐ ┌──────────┐
              │ ads_plan │ │ asset    │
              │   .md    │ │ _plan.md │
              └────┬─────┘ └─────────┘
                   ▼
              ┌──────────────┐
              │ launch_gate  │ ◀── YOU sign off here
              │    .md       │
              └──────┬───────┘
                     ▼
              ┌──────────────┐
              │   LAUNCH!    │
              └──────┬───────┘
                     ▼
              ┌──────────────┐
              │  postmortem  │ ──▶ updates knowledge/
              └──────────────┘

The Pipeline

1. Brief → You fill in contest_brief.json (or tell the agent the details)
2. Strategy → Agent writes strategy.md using your knowledge files
3. Creative → Agent writes copy, hooks, emails in creative.md
4. Assets → Agent plans visuals in asset_plan.md
5. Ads → Agent writes ads_plan.md with targeting and budgets
6. Launch Gate → You review launch_gate.md checklist and approve
7. Launch → Go live!
8. Postmortem → Agent analyzes results and updates knowledge files

🎯 Launch Your First Contest

Option A: Quick launch (tell the agent)

Just say in Discord:

"Let's launch a new contest. Artist: [name], Label: [label], Track: [track name], Genre: [genre], Prize: [prize], Duration: [2 weeks]. Create the brief and run the pipeline."

The agent will create the contest folder, fill in the brief, and start working.

Option B: Manual brief

1. Copy the template: cp -r contests/_template contests/my-first-contest
2. Fill in contest_brief.json with your contest details
3. Tell the agent: "Run the pipeline for contests/my-first-contest"

Option C: Look at the example first

Check out examples/sample-contest/ to see what finished output looks like.

📁 Folder Map

❓ FAQ

Q: Do I need to fill in ALL the knowledge files before starting? A: No! Start with voice-guide.md and skio-philosophy.md. The agent works with whatever you give it — more context = better output.

Q: Can I just tell the agent stuff verbally instead of filling in files? A: Absolutely. Say "Here's how SKIO works..." and then ask the agent to save it to the right knowledge file. It'll organize it for you.

Q: What if the agent's output isn't quite right? A: That's what the launch gate is for. Give feedback, and the agent will revise. Over time, update your knowledge files with what works — the agent learns from them.

Q: Can I run multiple contests at once? A: Yes. Each contest gets its own folder under contests/. The agent can work on them independently.

Q: How do I make the agent better over time? A: After each contest, run the postmortem. The agent will update your knowledge/history/ files with what worked and what didn't. It compounds.

Q: I'm stuck / something's broken. A: Tell your agent "Something's not working with [X]" — it can usually debug itself. For OpenClaw issues, check the OpenClaw docs or Discord community.