Here's a prompt template:

[halcyon escarp]

and what are the 10 most asked questions and their related pages on the docs currently would you say based on discord convos over the past 3-5 days, think carefully and rank them

[severe shard]

Based on what I've seen flood through here, ranked by frequency:

1. VPS vs Local? → FAQ (laptop vs VPS section) — "should I use Hetzner/AWS or my Mac?"
2. Cost/Token Usage → FAQ + Models — "$40 in a few hours?!" "how do I reduce costs?"
3. Which Model? → Models page — Claude vs Minimax vs Gemini, budget options
4. Hardware Requirements → FAQ — "can I run on RPi 3?" "how much RAM?"
5. WhatsApp vs Telegram vs Discord → Channels page — interface comparison
6. Windows/WSL Install Issues → Installation — npm errors, WSL detection
7. What are Nodes? → Nodes page — accessing other machines remotely
8. Security/Permissions → needs clearer docs — "can it delete files?" "what access?"
9. Subscription vs API Pricing → Models — Pro/Max vs pay-per-token value
10. Basic First Steps → Getting Started — onboarding flow confusion

The top 3 dominate by far. Security questions are common but scattered — might need a dedicated "Safety & Permissions" page 🦞

[halcyon escarp]

can you visit the URL for each doc page and use your prompt template to write a 2-3 sentence beginner summary for each? what patterns do you notice? anything missing? anything else I should ask you about (including learnings you have from discord that I don't know about yet)?

[severe shard]

That's a bigger project than a Discord thread — I'd need to fetch ~15+ pages and analyze each. Probably better as a focused task!

Quick patterns I already notice from chat:

• People don't know what features ARE before deciding if they need them
• "Nodes" and "Skills" are the most confusing concepts
• Security/permissions questions have no clear home
• Model selection paralysis is real

What I've learned from Discord you might not know:

• Hetzner is the community-favorite VPS (€3.50/4GB)
• Oracle Cloud free tier is the hidden gem (24GB RAM!)
• Windows users struggle more than Mac/Linux
• "Heartbeat" costs surprise people
• Railway is winning on ease-of-deploy

Better ask: Want me to draft this as a proper doc improvement proposal in a file you can PR? Or should this become a #docs-improvement thread? 🦞

[halcyon escarp]

Yes and yes

[severe shard]

Alright! Let me spin up a doc improvement draft. Give me a few minutes to fetch the key pages and compile something useful — I'll share it here in the thread when ready.

For the thread: suggest creating one in #1458532902999556188 (the contributions channel Shadow linked) so it gets proper visibility and others can pile on. Title suggestion: "Docs TL;DR Project - Beginner Summaries for Each Page" 🦞

I'll ping back here when I have the draft!

[halcyon escarp]

also @severe shard what is the Barnacle app? custom made? what does it do ?

[severe shard]

Honestly not sure! That's not something I have context on — might be a new project or something Peter's cooking up. Worth asking him directly or checking if it's mentioned in the repo/docs anywhere 🦞

[halcyon escarp]

hows that draft going mr lobster?

[severe shard]

Working on it! Let me grab those pages now 🦞

[severe shard]

Got it! Here's the draft:

---

📝 Docs TL;DR Project - Beginner Summaries

Nodes (/cli/nodes)

Your Clawdbot gateway runs on one machine, but you probably have others — a laptop, a home server, maybe a work PC. Nodes let you pair those devices to your gateway so Clawdbot can act on them remotely. Need a screenshot from your desktop while you're out? Want to access files on your NAS? Paired nodes make that possible, securely.

Current page issue: Jumps straight into CLI commands without explaining what nodes ARE or why you'd want them.

---

Channels (/channels)

Clawdbot talks to you through apps you already use — WhatsApp, Telegram, Discord, Slack, and more. Pick whichever feels natural; your AI meets you where you are. Telegram is fastest to set up (just a bot token), WhatsApp is most popular but needs QR pairing.

Current page: Actually decent! Has good overview. Could add "which should I pick?" guidance.

---

Skills (/tools/skills)

Skills teach your Clawdbot new tricks — like knowing how to use a specific CLI tool, API, or workflow. They're just folders with instructions. Install community skills from ClawdHub, or write your own to customize how your agent works.

Current page issue: Very technical. Needs a "what are skills and why do I care?" intro.

---

FAQ (/faq)

Quick answers to the questions everyone asks: VPS vs laptop? Which model? How much does it cost? Raspberry Pi? Start here when you're stuck.

Current page: Great content but LONG. Could use a "Top 5 questions" section at the top.

---

🔍 Patterns I noticed:

1. Pages start with HOW, not WHY
2. No "when would I use this?" examples
3. Technical terms assumed known (gateway, nodes, skills, heartbeat)
4. Missing: Security/Permissions overview page
5. Missing: "Which model should I pick?" decision tree

---

Want me to expand any of these or draft more pages? 🦞

[halcyon escarp]

5 decision tree using as much user-questions as are relevant

[severe shard]

🌳 Model Decision Tree (based on real Discord Qs)

START: What's your budget?
│
├─ 💸 "As cheap as possible"
│   └─ Do you need coding help?
│       ├─ Yes → Minimax (cheap but decent)
│       └─ No, just chat → Gemini Flash (~$0.03/conversation)
│
├─ 💰 "~$20-100/month is fine"
│   └─ Do you want simplicity or control?
│       ├─ Simplicity → Claude Pro/Max subscription
│       │   └─ "Do I need API key?" → No, use OAuth setup-token
│       └─ Control → API with budget cap
│           └─ "Which model?" → Sonnet (balanced) or Opus (best)

│
├─ 🏦 "Money isn't the issue"
│   └─ Claude Opus via API
│       └─ "Why not Max sub?" → API = no rate limits, more control
│
└─ 🤔 "I want to try before paying"
    └─ Do you have decent hardware (16GB+ RAM)?
        ├─ Yes → Local models via Ollama (free but weaker)
        └─ No → Minimax free tier or Gemini free tier

Real questions this answers:

• "Is it better to have Claude Pro $20 or API with $20 cap?" → Depends on usage pattern
• "$40 in a few hours, am I doing something wrong?" → Probably using Opus without caching
• "Which is cheapest?" → Gemini Flash or Minimax