This repo is a growth agent for the AI you already use.
Install these skills and Claude, ChatGPT, Grok, or Cursor learns how to interview you about your business, build your Growth Plan, watch the internet for people showing real buying intent, and draft your responses — all against your live Earth Care Network workspace.
Nothing ever sends without you. Every email, reply, and published page arrives as a draft that waits for your approval. That's not a setting — it's enforced by the platform.
You don't need to be technical to use this repo. Two terms, plainly:
A skill is a markdown file — a written playbook your AI reads and follows. There's no code in a skill and nothing runs on your computer. You can open every file in this repo and read exactly what your AI will be told to do. We encourage it — start with this one.
MCP is a plug. It's the standard that lets your AI use Earth Care Network on your behalf — reading your pipeline, drafting outreach, arming monitoring. You connect it with one sign-in screen (the same OAuth pattern as "Sign in with Google"). No API keys to paste, no passwords shared with the AI.
More terms as you meet them: reference glossary.
The skills walk you down one path. Each step unlocks the next:
Same session (about 20 minutes):
| Step | Time | What you have after |
|---|---|---|
| 1. Connect | ~2 min | Your AI can see your workspace |
| 2. Onboarding interview | 10–15 min | Your story, offer, and Growth Plan drafted from your own words |
| 3. Confirm your ICP | 2–3 passes | A definition of your ideal customer sharp enough to actually search for |
| 4. Arm a watchlist | same session | The internet is being monitored for your buyers' own words |
| 5. First prospect list | same session | Real companies matching your ICP, each with a reason it fits |
Ongoing (compounds weekly):
| Step | Time | What you have after |
|---|---|---|
| 6. Quality signals arrive | days | Real people showing buying intent, scored and triaged |
| 7. Drafted responses | minutes each | Replies and outreach written for you — you edit, you send |
Two honest expectations, because the skills are built to be honest with you:
- Your ICP is the hinge. Everything downstream — watchlist keywords, lead quality, outreach angles — inherits from it. Expect two or three tightening passes. That's the process working, not failing.
- Signals take days, not minutes. Monitoring compounds. The return is a growing pipeline of warm, evidence-backed conversations — not a burst of leads on day one.
For chat clients it's one OAuth sign-in — no API key to paste. Developer tools like Cursor use an API key you create in your workspace settings.
| Your AI | Where to connect |
|---|---|
| Claude (claude.ai) | Settings → Connectors → Add custom connector |
| ChatGPT | Settings → Connectors → Advanced → Developer mode |
| Grok | grok.com/connectors → New Connector → Custom |
| Cursor / scripts | API key — see the docs link below |
Server URL: https://www.earthcare.network/mcp/
Full per-client walkthroughs: earthcare.network/docs/mcp
Don't have an Earth Care Network account yet? The sign-in screen creates one — free to start.
Using claude.ai or ChatGPT in the browser? You may only need Step 1 — the connector alone gives your AI the tools, and you can paste any skill's text into the chat as instructions. Step 2 is for clients that read skill files from disk (Claude Code, Cursor, and similar).
Copy the contents of skills/ into wherever your AI reads skills from.
For Claude Code that's ~/.claude/skills/:
cp -R skills/* ~/.claude/skills/(That command copies every skill folder from this repo into Claude Code's
skills directory. -R means "include everything inside the folders.")
Each skill is a self-contained folder — install only the ones you want.
The skills trigger on their own. Try:
"Help me get set up with Earth Care Network" "Who should I be talking to this week?" "My leads are junk — fix my targeting" "Help me plan content that supports sales"
Eight skills in two arcs. They compose — onboarding hands off to ICP refinement when your targeting is thin; signals hand off to outreach drafting.
| Skill | What you have when it finishes | Say this to start |
|---|---|---|
| onboarding-with-earth-care-network | Full setup: your context, your enterprise, all seven Growth Plan modules, first leads | "Help me get set up with Earth Care Network" |
| refining-your-icp | A beachhead customer definition specific enough to name three real companies that fit | "My leads are junk — fix my targeting" |
| finding-buying-signals | Monitoring armed, and today's signals triaged into act / watch / drop with reasons | "Who should I be talking to this week?" |
| running-growth-experiments | A Growth Plan audit and the automations it recommends, armed as drafts | "Give me a growth-plan checkup" |
| enterprise-sales-context-interview | Durable sales context — voice, objections, response policy — so outreach stops sounding generic | "Interview me about my sales context" |
| Skill | What you have when it finishes | Say this to start |
|---|---|---|
| journal-strategic-planning | An editorial strategy grounded in your real audience, plus a six-week topic backlog | "Help me plan a content strategy" |
| journal-content-research | A source-backed research brief for one topic — verified facts, no invented claims | "Research this topic before we write it" |
| journal-content-writing | A drafted, reviewed article — published only after you explicitly approve | "Let's write that journal post" |
These rules aren't in the fine print. They're the product.
- Draft-first, always. Nothing outbound — email, reply, invitation, published page, automation — goes out without your explicit instruction. Outreach uses two separate gates: you approve a draft, then you (or your explicit "send it") trigger the send as a second step. The platform enforces this server-side; a skill can't override it even if it wanted to.
- Read before you install. Every skill is plain markdown. If a skill repo ever asks you to run a script you can't read, or to paste a password into a chat, walk away. This one never will.
- Metering is transparent. AI usage is metered to your workspace and the skills tell you before a costly discovery run. (If you self-host the platform with your own AI keys, you pay Earth Care Network nothing for AI — a platform feature, separate from these skills.)
- Nothing is invented. If only four leads matched, you get four. Skills that pad results to look productive are worse than useless — you'd act on them.
Found a security problem? See SECURITY.md.
No ecn tools available — the connector isn't attached. Redo Install
Step 1 and confirm the sign-in completed.
"Confirm your ideal customer first" — lead discovery unlocks after your
ICP is confirmed. Ask your AI to finish the Growth Plan through the icp
module.
A skill doesn't trigger — name it directly: "use the onboarding-with-earth-care-network skill".
Leads look wrong — your ICP is too broad. Ask your AI to refine your ICP, then re-run discovery. Two or three passes is normal.
Optional reading, whenever you're curious:
- What is Git, in plain language — the version-control system this repo lives in
- How agent skills work — the format these files follow
- How MCP works — the open standard connecting your AI to real tools
Want to improve a skill or add one? See CONTRIBUTING.md.
MIT — copy, adapt, and share these skills freely. The Earth Care Network platform itself is separate software (AGPL-3.0 core). See LICENSE.