Writing purpose, setting up GitHub, committing files
Claude
You instruct — Claude executes autonomously
Writing code, generating CHANGELOG, refactoring
Both
You initiate, Claude executes, you review
Reviewing code before committing, testing features
Auto
Triggered automatically — no action needed
CLAUDE.md loaded on init, Pages deploys on push
The 8 Core Principles
Principle
What it means in practice
When it fires
Security
Never expose keys. Use .env. Run security reviewer before every deploy.
Any time credentials or user data are involved
Low Token
One task per session. /compact when context grows. No padding.
Every session — enforce from message 1
Max Efficiency
Reuse prompts. Full file context. One task only.
Session start and task scoping
Max Memory
Update CLAUDE.md + memory.md every session without fail.
Session end — non-negotiable
Self-Learn
After every solution ask Claude to explain why. Absorb it. Note it.
After every working piece of output
Self-Execute
Give Claude full context and let it run. Review output. Don't micromanage.
Complex builds and repetitive tasks
Code Recheck
Every output tested before commit. Tester sub-agent on every deploy.
Before every commit and every deployment
Teach Human
Always ask 'why' not just 'what'. Claude explains. You learn.
After any fix, build, or explanation
Quick Commands Reference
# Claude Code / Antigravity essentials claude init # initialise Claude Code in project folder /status # see what Claude knows — files, CLAUDE.md, memory /memory add [fact] # save fact to persistent cross-session memory /memory list # view all saved memory /compact # compress context when it grows long /clear # clear conversation (keeps memory) /agents # manage sub-agents# Git essentials git status # what files changed git log --oneline -20 # last 20 commits with hashes git add . # stage all changes git commit -m "message" # commit with message git push # push to GitHub git checkout [hash] -- file # restore single file to specific commit git checkout -b branch-name # create + switch to new branch git stash # park uncommitted changes git stash pop # restore parked changes git revert [hash] # safe undo — creates new commit
Project Phases
Follow in sequence for every new project. Highlighted rows are critical — do not skip.
Skills (SKILL.md) are the evolved form of CLAUDE.md — reusable instructions that trigger automatically. Watch before building your first skill.
File
Purpose
Key Contents
Who
When
CLAUDE.md
AI instruction file
Stack, rules, status, session history
Both
Every session
CHANGELOG.md
Version history
Date, what changed, broke, fixed
Both
After every session
.env
Secrets — local only
Real API keys. NEVER commit.
You
Adding credentials
.env.example
Env template — public
Placeholder values only. Safe to commit.
You
Adding new env var
docs/memory.md
Cross-project learnings
Patterns, gotchas, reusable solutions
Both
After solving new problem
docs/prompts.md
Project prompt library
Reusable prompts tuned for this project
Both
When prompt worth saving
# CLAUDE.md template — fill every section # Project: [NAME]
# Last updated: [DATE]
# Status: [ACTIVE / PAUSED / BROKEN / COMPLETE] ## Purpose
[One sentence: what this does and who for] ## Tech Stack
- Frontend: [HTML/CSS/JS or React]
- Backend: [Supabase / None]
- Hosting: [GitHub Pages / Vercel]
- APIs: [list each API] ## Coding Rules
- Complete files only — no snippets - Explain every change in plain English (I am non-developer) - Flag before touching anything outside task scope ## Do Not Touch
- [list files/sections] ## Current Status
- Works: [list]
- Broken: [list]
- Next: [priority ordered list] ## Session History
- [DATE]: [what was done, outcome]
Claude Code in Action — Adding + Controlling context + Custom commands
Context management is the core skill of Claude Code. This teaches the exact session workflow this phase is built on.
Important
One session = one task. If you think of a second thing mid-session — open a new chat. Context dilution is real: more scope = worse output across everything.
☐
Task
Detail
Who
Principle
Line 1: project name + purpose
Never assume Claude remembers. Always restate. Every single chat.
You
Low Token
ONE task this session
Not "build everything". One feature, one fix, one file.
You
Max Efficiency
Paste complete relevant file
Claude works on what you paste — not what it imagines.
You
Low Token
Paste exact error text if debugging
The exact error string — not your description of it
You
Code Recheck
Specify output format
Complete file / function only / explanation / step-by-step
You
Low Token
After output: ask Claude to explain why
"Explain what you changed in 3 bullets and why each was needed"
Covers full GitHub integration with Claude Code — automated commits, PR creation, and connecting Claude Code to your repo.
Never
git reset --hard destroys uncommitted changes permanently. Always use git revert instead — it creates a new commit that undoes the change, keeping history intact.
☐
Task
Detail
Who
Principle
Commit before ANY change
Label: "snapshot before [change]". This is your restore point.
You
Security
Test in browser before committing
F12 → Console → zero red errors. Mobile too.
You
Code Recheck
Run Tester sub-agent before commit
Paste into QA chat. Action every CRITICAL finding.
Claude
Code Recheck
Commit message: [verb] [what] — [result]
"add contact form — tested working" not "update"
You
Max Memory
Update CHANGELOG.md
Claude can write this from your description
Both
Max Memory
Use branches for experiments
git checkout -b experiment/feature — never break main
You
Security
Claude Code in Action — MCP servers (browser testing)
Claude has no memory between chats by default. This is your architecture for continuity. All four layers are required.
Layer
What it is
Lives in
Lifetime
Who
Short-term
Current chat context — everything said so far
Conversation window
Until chat ends
Auto
Long-term
CLAUDE.md — project rules, stack, status
Repo root
Permanent until edited
Both
System
/memory — Claude Code persistent store
Claude Code internal
Persists across sessions
Both
Cross-project
memory.md — learnings across all projects
docs/memory.md in repo
Permanent, grows over time
Both
Session Transfer — How to Move to a New Chat
This is the most important memory skill. When a session gets long or you need to start fresh, this is how you transfer everything — context, decisions, status, next steps — without losing anything.
◈
The Session Transfer Protocol
1
End of current session — run the capture prompt
Before closing the chat, run the "End-of-session capture" prompt (see Prompts tab). Claude will produce a structured summary you can paste into the next chat.
2
Copy the full structured output
Claude will give you: current state, what was decided, what was built, what broke, what's next, and /memory commands to run. Copy all of it.
3
In Antigravity — run /memory commands
Paste and run each /memory add command Claude gave you. This saves the session facts to Claude Code's persistent memory.
4
Update CLAUDE.md — Status + Session History
Paste Claude's CLAUDE.md update suggestions. Commit the file. This is your long-term memory layer.
5
Open new chat — use the handoff prompt
Paste the full structured handoff at the top of the new chat. Claude reads it and continues exactly where you left off.
◈End-of-session capture — run before closing every chatMemory
Before we close this session, produce a structured handoff document: ## SESSION HANDOFF — [DATE]### Project State - Project: [name + one-sentence purpose]
- Stack: [current tech stack]
- Live URL: [if deployed]
- Repo: [GitHub URL] ### What We Completed This Session [List each thing built, fixed, or decided — numbered] ### What We Decided (with reasons) [List key architectural or design decisions made and WHY] ### Current Working State - Works: [list what currently functions correctly]
- Broken: [list what is currently broken with suspected cause]
- Untested: [list what was built but not yet verified] ### Next Session — Priority Order 1. [highest priority task]
2. [second task]
3. [third task] ### /memory Commands to Run Now [List each: /memory add [specific reusable fact]] ### CLAUDE.md Updates Needed [Exact text to paste into Status and Session History sections] ### memory.md Additions [Bullet points under correct headings, max 15 words each] ### Warnings for Next Session [Any gotchas, fragile code, things to be careful of] Keep every section. Leave nothing blank. This document will be pasted into the next chat as the opening message.
↻New chat handoff opener — paste at top of every new sessionMemory
I am continuing work on a project. Here is my full session handoff: [PASTE THE ENTIRE HANDOFF DOCUMENT FROM THE PREVIOUS SESSION] Rules for this session:
- You have full context from the handoff above — do not re-ask what I already told you
- Complete files only — no snippets
- Explain every change in plain English (I am a non-developer)
- Flag before touching anything outside the task scope
- One task at a time unless I explicitly ask for more Today's task: [ONE specific task from the priority list above] Current file to work on: [PASTE FULL FILE CONTENT]
Memory Layer How-To
◈ SHORT-TERM · Trigger: mid-session discovery
How
Nothing — stay in the same chat. Claude holds everything in the active window automatically.
Who
Auto — no action needed
Limit
~200K tokens. When context gets long, run /compact — Claude summarises and frees space.
◈ SYSTEM / /memory · Trigger: learned something that applies to ALL future projects
How
Type in Antigravity: /memory add [fact]. Immediate and persists across all sessions.
Who
You — type the command. Also in chat: "Add to memory: [fact]"
Examples
/memory add Supabase RLS must be enabled or all reads return empty
/memory add Always commit before making any changes to working code
◈ CROSS-PROJECT / memory.md · Trigger: end of every productive session
How
Ask Claude to draft additions. You paste into docs/memory.md and commit.
Who
Claude drafts. You paste and commit. Both responsible.
Prompt
Summarise what we learned today for memory.md as bullets under the correct heading. Max 15 words each. Only reusable cross-project knowledge.
Prompt Library
Copy-paste ready. Add your own to docs/prompts.md in each project.
⬡Standard session openerSession
Project: [NAME] Purpose: [one sentence] Stack: [e.g. HTML/CSS/JS + Supabase + Vercel] Current state: [working / broken at X / starting fresh] My level: Non-developer. I build with Claude Code + Antigravity. This session task: [ONE specific task only] Output required:
- Complete updated file (not a snippet)
- Exact file name and where to place the code
- Do not change anything outside this task scope
- Flag anything that could break existing functionality
- After the code: explain in plain English what changed and why Current file: [PASTE ENTIRE FILE CONTENT HERE]
◈End-of-session memory captureMemory
Before we close this session, produce a structured handoff: ## SESSION HANDOFF — [DATE] ### Project State
- Project: [name + purpose] - Stack: [current stack] ### What We Completed [numbered list] ### Decisions Made (with reasons) [key decisions + why] ### Current State
- Works: [list] - Broken: [list with suspected cause] ### Next Session Priority Order
1. [highest priority] 2. [second] ### /memory Commands to Run [each: /memory add [fact]] ### CLAUDE.md Updates Needed [exact text for Status + Session History] ### memory.md Additions [bullets under correct headings, max 15 words each] ### Warnings for Next Session [gotchas, fragile code, things to be careful of]
↻New chat handoff openerMemory
Continuing work. Full session handoff: [PASTE ENTIRE HANDOFF DOCUMENT] Rules: Complete files only. Plain English explanations. One task. Flag before touching anything outside scope. Today's task: [task from priority list] Current file: [PASTE FULL FILE]
⬡Add a feature without breaking existing codeBuild
Add [feature] to this file. Give me the complete updated file.
Do not remove or change anything not related to this task.
Flag anything that could break existing functionality.
After the code: explain in plain English what changed and why. Current file: [PASTE ENTIRE FILE]
⬡Build from scratch — new componentBuild
Build a [component/page/feature] with these requirements:
- [requirement 1] - [requirement 2] - [requirement 3] Stack: [HTML/CSS/JS / React / etc] Output: Complete file, ready to use. No placeholders.
After the code: explain the key parts in plain English.
⬡Architecture review before buildingBuild
I am planning: [describe what you want to build] My planned approach: [how you think you'd build it] Stack I'm considering: [list] Review this plan. Flag:
- Anything that will cause problems
- Simpler alternatives if my approach is overcomplicated
- Security considerations I'm missing
- The one most important thing to get right first I am a non-developer building with AI tools. Be direct.
⬡Fix a bugDebug
This code has a bug.
Error: [PASTE EXACT ERROR TEXT FROM CONSOLE] Expected: [what should happen] Actual: [what is happening] Tell me: exact line to change + what to change it to.
Then explain why that line caused the error — I want to understand it. File: [PASTE FULL FILE]
⬡Explain this codeDebug
Explain what this code does in plain English. I am a non-developer.
1. What does each major section do? (one sentence each)
2. What would break if I changed [specific part]?
3. Is there anything in here that looks risky or fragile?
4. What is the most important concept to understand here? Code: [PASTE CODE]
⬡Why is this not working (general)Debug
This is not working as expected.
What I did: [describe steps] What I expected: [expected outcome] What happened: [actual outcome] Console errors (if any): [paste exact text or "none"] Diagnose the most likely causes in order of probability.
For each: tell me how to confirm it is the cause, then how to fix it. Relevant files: [PASTE FILES]
⬡Tester sub-agent — run before every commitReview
You are a senior QA engineer. I am about to commit this code.
Find bugs, edge cases, and things that will break in production.
Exact line number + what fails + why. Do not be encouraging. Also check:
- console.log statements left in (remove before deploy)
- hardcoded values that should be in .env
- functions that fail on empty or null inputs
- breaks at screen width under 768px
- any API calls without error handling Report: CRITICAL (must fix) / WARNING (should fix) / NOTE (optional) Code: [PASTE ALL CHANGED FILES]
⬡Security reviewer — before every deployReview
You are a senior security engineer. Review for vulnerabilities.
Flag if present:
- API keys, tokens, or secrets visible in frontend
- Unvalidated or unsanitised user inputs
- Data accessible without auth that requires it
- CORS misconfigurations
- XSS attack vectors
- Sensitive data logged to console Severity: Critical / High / Medium. Line number. How to fix. Code: [PASTE CODE]
⬡Final pre-deploy review — all changed filesDeploy
I am deploying to production. Complete review of all changed files. SECURITY: API keys in frontend / unvalidated inputs / auth gaps
CODE QUALITY: console.log remaining / hardcoded .env values / null input failures
MOBILE: breaks at <768px / images without dimensions
PERFORMANCE: obvious slow operations / redundant API calls Report: CRITICAL (block deploy) / WARNING (fix soon) / NOTE (optional) Files: [PASTE ALL CHANGED FILES]
⬡Teach me what I just builtLearn
I just built [feature]. Teach me:
1. What does each major section do? (1 sentence, plain English)
2. What would break if I changed [specific part]? Why?
3. What is the one concept I most need to understand here?
4. What would a senior developer do differently, and why?
5. What should I add to memory.md from this? Code: [PASTE]
⬡Explain a concept I don't understandLearn
Explain [concept] to me simply. I am a non-developer.
1. One real-world analogy (not software)
2. One practical example from web development
3. What is the one thing I need to be able to DO with this?
4. What breaks if I get this wrong?
5. Should I add anything about this to memory.md?
⬡Architect sub-agent — review a design decisionLearn
You are a senior software architect. Review this decision.
Decision: [describe what you are planning to build or how] Tell me:
1. What is wrong or risky about this approach?
2. What would you do differently and why?
3. What are the 3 most important things to get right?
4. What will cause the most pain at scale? I am a non-developer. Explain trade-offs simply.
Write a CHANGELOG.md entry for these changes.
Format: ## [version] - [date] / ### Added / ### Fixed / ### Changed / ### Known Issues
Keep each bullet under 15 words. Plain English. Changes made this session: [describe or paste diff]
⬡Refactor — improve without breakingBuild
Refactor this code to be cleaner and more maintainable.
Rules:
- Do NOT change any functionality — same inputs, same outputs
- Explain every change you make and why it improves the code
- Flag anything you are unsure about before changing it
- Give me the complete refactored file Code: [PASTE]
⬡Generate .env.example from codeDeploy
Scan this code and identify every environment variable being used.
Generate a .env.example file with:
- Every variable name found
- A clear comment explaining what each one is for
- Placeholder values (never real values)
- Group related variables together Code: [PASTE ALL PROJECT FILES]
⬡Scope check — is my task too big for one session?Session
I want to do this in one session: [describe full task] Is this too much for one session?
If yes: break it into the minimum number of focused sessions.
For each session: state the single task, estimated complexity, and what the output will be.
Order them by dependency — what must be done first.
Skills & Tools
Everything you need to know about, how to get it, and when to use it. Must = need for almost every project. Should = most projects. Nice = specific use cases.
Core Stack
MCP Tools
Design Skills
Data & APIs
AI Tools
Skills Matrix
🤖
Claude Code / Antigravity
AI Coding Agent · Must
Reads your files, writes code, runs terminal commands. Your primary build tool. Use for all code generation, refactoring, and debugging.
Claude working directly with your local files, folders, apps. File tasks, research, Excel/PowerPoint, scheduled automation. Your non-code workflow engine.
No-code workflow automation. Connect apps, trigger actions, run on schedule. Self-host free or cloud paid. Better than Zapier for complex flows. Essential for agentic pipelines.
MCP (Model Context Protocol) tools extend what Claude Code can do. Install them via claude mcp add [name]. Each gives Claude a new capability.
# Install MCP tools in Antigravity claude mcp add github # read/write repos, create PRs, manage issues claude mcp add supabase # query DB, manage tables, check RLS from Claude claude mcp add memory # persistent memory outside conversation window claude mcp add puppeteer # browser automation — Claude opens URLs, clicks, tests claude mcp add filesystem # read/write local files (built-in to Claude Code)
🔥
Firecrawl MCP
Web Scraping · Should
Let Claude scrape any website and extract structured data. Use for competitor research, building directories, monitoring pages, data collection without APIs.
Claude controls a real browser — navigates, clicks, fills forms, takes screenshots. Use for testing your live site, automating data entry, monitoring changes.
Query tables, inspect RLS policies, check auth users — all from Claude Code without going to the Supabase dashboard. Massive time saver for debugging data issues.
Design skills are the set of prompts, references, and rules you give Claude to produce consistently good-looking output. You don't need to be a designer — you need the right instructions.
🎨
Frontend Design Skill
SKILL.md · Must for any UI
A SKILL.md file that instructs Claude on typography, colour systems, layout principles, and how to avoid generic AI aesthetics. Auto-triggers on any UI task.
Free, high-quality web fonts. Always specify a font stack in your CLAUDE.md so every component uses the same typeface. Avoid Inter and Roboto — overused.
Utility-first CSS framework. Claude knows it extremely well — specify "use Tailwind" and output is consistently styled. Use CDN version for quick builds.
Clean, consistent SVG icons. Available as React components or vanilla SVG. Claude knows the full icon set — just say "use Lucide icons" in your prompt.
⬡Build a page with design system constraintsDesign
Build a [page/component] that matches this design system:
- Font: [font name from Google Fonts] - Primary colour: [hex] - Background: [dark/light, hex] - Style: [minimal/editorial/dashboard/bold/soft] - Framework: [Tailwind / vanilla CSS / shadcn] - Icons: Lucide Avoid: generic AI aesthetics, Inter font, purple gradient backgrounds.
Make it look like it was professionally designed, not AI-generated. Content needed: [describe the content/sections]
📡
REST APIs (general)
Data · Must understand
All external services talk via REST API. Learn: what a GET/POST request is, how to pass headers, what JSON is. Claude will write the fetch() calls — you need to understand the shape of the data.
Industry standard for payments. Claude knows the Stripe API well. Use for subscriptions, one-time payments, checkout flows. Test mode lets you build without real money.
Build AI features directly into your apps. Claude writes the API calls. Essential for: chatbots, content generation, classification, summarisation. Learn after you can read JS.
n8n has built-in AI agent nodes. Connect Claude to any tool (email, Sheets, Slack, CRM) in an automated workflow without code. The non-developer's agentic stack.
Store embeddings for semantic search. Use when you need "search by meaning not keyword" — knowledge bases, recommendation systems, document Q&A. pgvector works inside Supabase.
Use this to audit your gaps before starting a new project. Must = you need this to ship anything. Should = most projects need it. Nice = specific use cases only.
Skill
Level
How to learn
Why you need it
Read a JS console error
Must
F12 → Console → Google the exact red text
Cannot debug without this. Everything else depends on it.
Edit a file in GitHub web UI
Must
Repo → file → pencil → edit → commit
Your fallback when Antigravity isn't available
What an API key is + .env pattern
Must
See Security tab in Principles section
Every project uses at least one API. Must not expose it.
git commit + git push (CLI)
Must
Claude Code in Action course, Phase 6
Web UI is fine but CLI is faster and more reliable
claude init + /status + /memory
Must
Claude Code in Action course
Foundation of every Claude Code session
JSON — what it looks like, how to read it
Must
Ask Claude: "Explain JSON with 3 examples"
All APIs return JSON. You must understand the shape of data.
HTML/CSS basics — div, class, id, flexbox
Should
freeCodeCamp Responsive Web Design (free)
Lets you read and modify Claude's frontend output
Async/await — what it means
Should
Ask Claude: "Explain async/await with an analogy"
All API calls are async. You'll see this in every JS file.
Supabase RLS — what it does
Should
Supabase docs + /memory add the gotcha
Skipping RLS = your database is publicly readable
Git branch + merge + conflict resolve
Should
Claude Code in Action + practice
Needed the moment you and your son work on the same repo
Vercel env vars + deployment config
Should
Vercel docs · 10 min read
Your .env values need to live somewhere in production
MCP tool installation
Should
Intro to MCP course
MCPs multiply Claude's capabilities — learn after basics
Prompt engineering patterns
Should
Anthropic prompting docs at docs.claude.com
Better prompts = better output = fewer iterations
React fundamentals
Nice
React.dev official tutorial
Needed for complex UIs with lots of state
Python basics
Nice
freeCodeCamp Python · 6 hrs
Many AI/data tools use Python. Claude writes it — you need to read it.
Project Types
Tool stacks, required skills, and starter prompts for the most common startup project types.
📊
Data Dashboard
Internal tools, analytics, tracking, reporting
Core Stack
HTML/CSS/JSChart.jsSupabaseVercelReact (if complex)Google Sheets API
Key Skills Needed
JSON parsingREST API fetchSupabase queriesAsync/awaitChart.js config
⬡Data dashboard starter prompt
Build a data dashboard with these specs:
Data source: [Supabase table / Google Sheets / static JSON / API endpoint] Metrics to show: [list each metric] Chart types: [line / bar / doughnut / mixed] Refresh: [real-time / on load / manual] Auth required: [yes - Supabase / no - public] Stack: HTML/CSS/JS + Chart.js. No frameworks unless necessary.
Style: dark theme, minimal, professional. Use [font name].
Build a SaaS app skeleton with:
- Auth: Supabase email+password login, signup, logout
- Protected routes: redirect to login if not authenticated
- Pricing tiers: [free / pro / enterprise — list what each includes] - Billing: Stripe checkout for paid tier, webhook to update user role in Supabase
- Dashboard: shows user's current plan and usage of [main feature] Stack: React + Supabase + Stripe + Vercel.
This is an MVP skeleton — no extra features yet.
15 courses reviewed. These are the ones that matter for your level. Do NOW first, NEXT second, skip the rest.
Honest assessment
You are past beginner AI literacy. You are not yet a developer. Your zone is tools, workflow, memory, sub-agents, and skills. That is this learning path.
Both bookmark it. When you update and commit, it live-updates for both.
Option 2 — Claude Project (per-person)
claude.ai → Projects → New project → Upload AI_Project_Playbook_v4.docx as knowledge
Add instruction: "You are a senior AI-native builder assistant. Always refer to the playbook. Both users are non-developers building with AI tools — explain everything simply."
Note
Claude Projects are per-account. Each person creates their own and uploads the same docx. You cannot share a single Project between two accounts currently.
Collaboration Rules (Shared Repo)
Rule
Why
Enforced by
Same CLAUDE.md structure on every project
Either person can pick up the other's project without confusion
Both
Same commit message formula
Shared history is readable by both
Both
Never commit to main during active collab
Use branches: feature/[name]. Merge only when tested.
Both
One person owns each task at a time
Prevents file conflicts and overwrites
Both
CHANGELOG.md updated after every session
Other person knows exactly what state the project is in
Both
Tester sub-agent before any merge to main
Never merge broken code
Whoever wrote the code
memory.md is shared in the repo
Both accumulate and benefit from shared learnings
Both — commit after every session
SKILL.md Setup — Reusable Across Projects
What is a Skill?
A SKILL.md is a markdown file Claude Code reads automatically when it detects a matching task. Write once — Claude applies it every time without repeating yourself. Watch Introduction to Agent Skills first.
---name: project-init description: Use when starting a new project, initialising a repo,
setting up CLAUDE.md, or asking how to structure a new build ---# Project Init — follow in sequence## Phase 1 — Define first (before any tool) - Project name: lowercase-hyphenated
- Purpose: [Tool] that helps [who] to [outcome]
- MVP features: maximum 5, numbered
- Out-of-scope: "v1 will NOT include..." ## Phase 2 — Repo - Create GitHub repo with README + .gitignore
- .gitignore BEFORE .env — always
- Enable GitHub Pages ## Session rules (enforce every session) - Complete files only — no snippets
- Plain English explanations (non-developer)
- Flag before touching anything outside task scope
- One task per session ## Security rules - Never put API keys in code — use .env
- Run security review before every deploy
- Grep for sk-, secret, password before committing
Digital Marketing
Every role you carry — Website, SEO, Content, Social, Ads, Analytics, Reputation, Brand — powered by AI. Tools, prompts, sub-agents, and automations for each.
Overview
SEO
Content
Social Media
Meta & Google Ads
Analytics
Brand & Reputation
Email Marketing
Video
Automations & n8n
Sub-Agents
Proposals
Your Advantage
You are a one-person marketing department. AI lets you operate like a team of 6. The goal: Claude handles the repetitive, research-heavy, and draft-production work so you focus on strategy, client relationships, and creative direction.
Your AI Marketing Stack — Master Map
Your Role
AI Tool
What AI does for you
Time saved
Website Design
Claude + Figma AI + Framer
Generate layouts, write copy, build pages with code
Brand Voice Skill in Claude Projects — set up before anything else
Claude 101 — Introduction to Projects
NOW
Content writing workflow with Claude + Surfer SEO
Surfer SEO YouTube channel
NEXT
Social media content calendar automation via n8n
n8n.io/workflows/categories/marketing
NEXT
Meta Ads analysis + weekly reporting workflow
n8n Meta Ads templates
NEXT
Google Analytics automated reporting with Claude insights
Looker Studio + n8n GA4 template
LATER
Full reputation monitoring pipeline
Brandwatch + n8n setup guides
SEO — AI-Powered Tools & Workflow
🔍
Semrush
All-in-one SEO · Must
Keyword research, site audit, competitor analysis, backlink tracking, local SEO, and now AI visibility tracking. The most comprehensive single SEO platform. Claude has a Semrush MCP — connect them.
Tells you exactly what a top-ranking page needs: word count, headings, keyword density, LSI terms. Write content in the Surfer editor and it scores in real-time. Claude can write the draft, you optimise in Surfer.
Best backlink analysis in the industry. AI Content Helper now included. Keyword Explorer shows what people actually search. Webmaster Tools free tier is useful for site owners.
Free. Shows exactly what queries drive traffic to your site, click-through rates, and indexing issues. Connect to Looker Studio + Claude for automated insight reports.
Let Claude scrape competitor pages and extract their heading structure, keyword usage, meta tags, and word count. Feed into your content strategy instantly.
Manage Google Business Profile, track local rankings, monitor and respond to reviews, manage citations across directories. Essential for any client with physical locations.
I need a keyword research strategy for:
Business: [business name + what they do] Location: [city/region if local] Target audience: [describe ideal customer] Competitors: [list 2-3 competitor URLs] Produce:
1. 10 primary target keywords (high intent, achievable difficulty)
2. 20 long-tail keywords grouped by topic cluster
3. 5 featured snippet opportunities (question-based queries)
4. Local SEO keyword variations if location-based
5. Content gap opportunities competitors are ranking for Format as a table: keyword | monthly volume estimate | intent | priority (H/M/L)
⬡SEO content brief generatorSEO
Create a detailed SEO content brief for this article:
Target keyword: [primary keyword] Supporting keywords: [3-5 secondary keywords] Search intent: [informational / transactional / navigational] Target audience: [describe] Competitor URLs to outrank: [paste 3 URLs] Brief must include:
- Recommended title (H1) — 3 variations
- Recommended meta description — 2 variations (under 155 chars)
- Exact heading structure (H2s and H3s)
- Word count recommendation
- Must-include topics and subtopics
- Internal linking suggestions
- Schema markup type to use
- CTA recommendation
⬡Competitor content gap analysisSEO
Analyse these competitor pages and identify content gaps I should fill:
My site: [your URL] Competitor 1: [URL] Competitor 2: [URL] My target keyword: [keyword] From the content at these URLs (or describe what you know about them):
1. What topics do they cover that I don't?
2. What questions do they answer that I'm missing?
3. Where is my content weaker (depth, structure, examples)?
4. What unique angle can I take to differentiate?
5. Recommend 5 new content pieces I should prioritise
⬡Technical SEO audit checklistSEO
Produce a technical SEO audit checklist for:
Website: [URL] Platform: [WordPress / Webflow / custom / etc] Categorise by: Critical (fix immediately) / Important (fix this month) / Nice-to-have Include checks for: page speed, Core Web Vitals, mobile responsiveness, crawlability, sitemap, robots.txt, canonical tags, structured data/schema, internal linking, broken links, image alt text, meta tags, heading hierarchy, HTTPS, duplicate content
Content Writing — AI Workflow
The Pro Workflow
Claude writes the first draft in your brand voice → Surfer SEO scores it → you refine the human angle → publish. This cuts writing time by 70% without losing quality.
✍
Claude (Projects + Skills)
Primary Writer · Must
Set up a Claude Project with your brand voice, target audience, tone examples, and banned phrases. Every content piece stays on-brand without re-briefing. Add a Content Skill that auto-loads your writing rules.
Final polish on every piece of content. Catches tone inconsistencies, readability issues, and grammar errors. The business tier adds brand tone guidelines. Install the Chrome extension — it works everywhere.
Manage your content calendar, briefs, and drafts in one place with AI summaries, rewriting, and translation built in. Share with clients for feedback and approvals.
Generate blog hero images, social post visuals, and ad creatives from text prompts. DALL-E 3 is accessible via ChatGPT Plus. Midjourney produces more photorealistic results.
Write a complete SEO-optimised blog post: Title: [H1 title from your brief] Target keyword: [primary keyword — use naturally, not stuffed] Supporting keywords: [include these throughout] Word count: [target word count] Audience: [describe reader — level, pain points, goals] Brand voice: [professional/conversational/authoritative/friendly] Banned phrases: [any phrases to avoid — e.g. "game-changer", "leverage"] Structure:
- H1: include primary keyword
- Introduction: hook + what reader will learn
- Body: follow this heading structure: [paste H2s and H3s from brief]
- Conclusion: summarise + clear CTA
- Meta description: under 155 chars, include keyword, include a benefit After writing: flag any section that feels thin and needs more depth.
⬡Repurpose one piece into multiple formatsContent
Repurpose this content into 5 formats:
[PASTE ORIGINAL BLOG POST OR ARTICLE] Create:
1. LinkedIn post (150-200 words, hook opening, 3 key points, CTA)
2. Instagram caption (80-100 words, engaging question, 5 hashtags)
3. Twitter/X thread (5 tweets, numbered, hook in tweet 1)
4. Email newsletter intro (150 words — tease the full piece, drive click)
5. Short video script (60 seconds, spoken word, punchy sentences) Keep the core message consistent. Adapt tone for each platform.
⬡Brand voice setup for Claude ProjectContent
I am setting up a Claude Project for [CLIENT/BRAND].
Analyse these sample content pieces and extract the brand voice:
[PASTE 3-5 EXAMPLES OF THEIR EXISTING CONTENT] Produce a brand voice document with:
1. Voice characteristics (5 adjectives that describe the tone)
2. Do's — phrases, structures, approaches to use
3. Don'ts — things this brand never says or does
4. Target audience summary (who we're writing for)
5. Example paragraph rewrite showing the voice in action
6. 10 banned/overused phrases to always avoid This will become the brand voice instruction in the project.
Social Media — AI Tools & Workflows
🎨
Canva AI
Visual Creation · Must
AI-powered image generation, Magic Resize (one design → all formats), Brand Kit (logos, fonts, colours auto-applied), and bulk content creation. Magic Write for captions. The go-to for non-designers.
Schedule posts across all platforms. AI assistant suggests best posting times, generates captions from URLs or prompts. Free tier is generous. Integrates with n8n for automated scheduling workflows.
Best tool for Instagram visual planning — drag and drop calendar shows how your grid looks before posting. AI caption writer, hashtag suggestions, best time to post analysis.
AI chatbots for Instagram DMs, Facebook Messenger, WhatsApp. Automate comment replies, lead capture, product recommendations. Integrates with your CRM and email lists via n8n.
Create a 30-day social media content calendar for:
Brand: [brand name + what they do] Platforms: [Instagram / Facebook / LinkedIn / TikTok — list all] Posting frequency: [e.g. daily Instagram, 3x/week LinkedIn] Audience: [describe] Goals this month: [awareness / leads / engagement / sales] Key dates: [upcoming events, product launches, holidays] Brand voice: [tone description] For each post include:
- Date and platform
- Content type (carousel / reel / static / story)
- Caption (complete, ready to post)
- Hashtags (15 for Instagram, 3-5 for LinkedIn)
- Visual direction (describe what the image/video should show)
- CTA Format as a table. Export-ready.
⬡Instagram carousel scriptSocial
Write an Instagram carousel post:
Topic: [topic] Goal: [educate / entertain / sell / build trust] Slides: [6-10 slides] Brand voice: [tone] For each slide:
- Slide number
- Headline (under 8 words — bold, punchy)
- Body text (under 30 words — clear, scannable)
- Visual direction Slide 1: Hook — must make them swipe
Last slide: CTA + save prompt
Caption: 150 words + 5 relevant hashtags
⬡Comment reply templates for reputationSocial
Create 10 ready-to-use comment reply templates for:
Brand: [brand name] Industry: [industry] Templates for:
1. Positive review / compliment (2 variations)
2. Negative complaint (2 variations — empathetic, de-escalating)
3. Product/service question (2 variations)
4. Request for more info (1 variation)
5. Spam / irrelevant comment (1 variation — polite ignore)
6. Influencer / partnership inquiry (1 variation)
7. Crisis / sensitive situation (1 variation) Each template: natural-sounding, brand-aligned, not robotic. Include [PLACEHOLDER] tags where personalisation is needed.
Meta & Google Ads — AI Tools
🎯
AdCreative.ai
Ad Creative · Should
Generate high-converting ad creatives at scale. Input your brand, product, and target audience. Outputs complete ad sets with images, headlines, and copy. Integrates with Facebook and Google Ads directly.
Connect Claude directly to your Meta Ads account. Ask in plain English: "Which ad sets are underperforming?", "What's my CPL this week?", "Pause everything with ROAS under 2." Claude reads and acts.
Google's AI now runs Performance Max campaigns automatically. Your job: provide the best assets (headlines, descriptions, images, videos). Claude writes the copy variants. You supply the strategy.
Write 5 Meta Ad copy variations for:
Product/Service: [describe] Offer: [what's being promoted — discount, free trial, consultation, etc] Target audience: [age, interests, pain points] Campaign objective: [awareness / traffic / leads / conversions] Brand voice: [tone] For each variation write:
- Primary text (125 chars max — hook first)
- Headline (40 chars max)
- Description (25 chars max)
- CTA button: [Learn More / Shop Now / Sign Up / Contact Us] Variation approaches:
1. Pain-focused (lead with the problem)
2. Benefit-focused (lead with the transformation)
3. Social proof (lead with results/numbers)
4. Curiosity/question hook
5. Direct offer (lead with the deal)
⬡Ad performance analysisAds
Analyse this Meta Ads performance data and give me actionable recommendations:
[PASTE YOUR AD DATA — campaign name, spend, impressions, clicks, CTR, CPL, ROAS] I need:
1. Which campaigns/ad sets to scale (and by how much)
2. Which to pause or kill (and why)
3. Audience insights from the data
4. Creative fatigue indicators
5. Budget reallocation recommendation
6. Top 3 tests to run next week My monthly budget: [budget] Primary KPI: [CPL / ROAS / CPM / CTR] Target: [e.g. CPL under $25]
⬡Google Ads RSA headlines + descriptionsAds
Write Google Responsive Search Ad assets for:
Business: [name + what they do] Target keyword: [primary keyword being bid on] Landing page: [URL or describe the page] USPs: [3-5 unique selling points] Location: [if local] Produce:
- 15 headlines (under 30 chars each — include keyword in at least 5)
- 4 descriptions (under 90 chars each)
- 2 sitelink extensions with descriptions
- 4 callout extensions Follow Google's best practices: keyword in first headline, benefit in second, brand in third.
Analytics — AI-Powered Reporting
📈
Looker Studio (free)
Reporting · Must
Google's free dashboard builder. Connect GA4, Google Ads, Search Console, and Meta Ads into one visual report. Share with clients as a live link. Claude can write the analysis narrative.
Google Analytics 4 with BigQuery export (free) lets Claude query your raw event data in plain English. Ask: "What's my top traffic source for leads?" without being a SQL expert.
Pre-built n8n template: pull GA4 + Google Ads + Meta Ads data weekly, send to Claude for AI analysis, email the insight report to you and clients automatically every Monday.
Interpret this GA4 data and give me a client-ready narrative:
[PASTE GA4 EXPORT OR KEY METRICS] Period: [date range] Compare to: [previous period] Business goal: [e.g. lead generation / ecommerce / brand awareness] Produce:
1. Executive summary (3 sentences max — what happened, why it matters)
2. Top 3 wins this period (with specific numbers)
3. Top 3 concerns or anomalies (with suspected causes)
4. Channel performance breakdown (organic, paid, social, direct, referral)
5. Conversion funnel analysis (where are users dropping off?)
6. Recommended actions for next period (prioritised) Write in plain English for a non-technical client.
⬡Monthly client report narrativeAnalytics
Write a monthly marketing report narrative for a client:
Client: [business name + industry] Period: [month + year] Data:
- Website: [sessions, users, bounce rate, avg session duration] - SEO: [organic traffic, keyword rankings changes] - Social: [follower growth, engagement rate, reach, best performing post] - Ads: [spend, impressions, clicks, conversions, ROAS/CPL] - Email: [open rate, click rate, subscribers, unsubscribes] Write:
1. Month in review (1 paragraph — story of the month)
2. What worked (celebrate the wins with specifics)
3. What we're watching (honest about what needs work)
4. Next month priorities (3 clear actions)
5. One strategic recommendation Tone: professional, confident, client-friendly. Not corporate-jargon.
Brand & Reputation Management
👁
Brandwatch
Social Listening · Should
Monitor every mention of your brand or clients across social media, news, blogs, and forums. AI sentiment analysis. Set up alerts for negative sentiment spikes. Export data for Claude analysis.
Monitor and respond to Google My Business and other platform reviews from one dashboard. Set up alerts for new reviews. Claude drafts the responses — you approve and send.
Free email alerts whenever your brand, clients, or competitors are mentioned online. Simple to set up. Pipe alerts to n8n → Claude → Slack for automated reputation monitoring.
Write 3 variations of a Google review response for:
Business: [business name + industry] Review: [paste the review] Star rating: [5 / 4 stars] Requirements:
- Thank the reviewer by first name if mentioned
- Acknowledge the specific thing they praised
- Reinforce a brand value naturally
- Invite them back with a soft CTA
- Under 80 words each
- Sound human — not like a template
- No emojis unless the reviewer used them
⬡Negative review responseReputation
Write a professional response to this negative review:
Business: [business name] Review: [paste the review] Response requirements:
- Acknowledge the frustration (don't be defensive)
- Apologise for the experience (not admitting fault if disputed)
- Offer to resolve — move conversation offline with contact details
- Do NOT include anything that could be used against the business
- Under 100 words
- Professional but warm tone Also write: a private follow-up message to send directly to the reviewer to resolve the issue
⬡GMB post content calendarGMB
Create 8 Google My Business posts for:
Business: [name + what they do] Location: [city] Posting frequency: [weekly] For each post:
- Post type: What's New / Event / Offer / Product
- Title (under 58 chars)
- Body (150-300 words, include local keyword naturally)
- CTA button: [Book / Order / Learn More / Call / Sign Up] - Offer details (if applicable) Topics to cover: services, team, community, FAQ, tips, promotions
Email Marketing — AI Tools & Sequences
📧
Klaviyo
E-commerce Email · Should
Industry standard for e-commerce email. AI-powered send time optimisation, predictive segments, and flow automation. Direct Shopify/WooCommerce integration. Claude writes the sequences, Klaviyo delivers them.
Mailchimp for simpler campaigns with strong free tier. ActiveCampaign for advanced automation sequences and CRM integration. Both have AI subject line generators and send time optimisation.
Best-in-class newsletter platform. Free tier is generous. Developer API lets you add subscribers programmatically. Built-in monetisation with paid subscriptions and ad network. Great for client newsletters.
Write a 5-email welcome sequence for new subscribers:
Brand: [name + what they do] Audience: [who subscribes + why they signed up] Goal of sequence: [build trust / educate / convert to customer] Brand voice: [tone] Lead magnet delivered: [what they signed up for] Email 1 (immediate): Deliver the lead magnet + warm welcome
Email 2 (day 2): Your brand story — why you do what you do
Email 3 (day 4): Biggest insight or value piece (educate)
Email 4 (day 6): Social proof — results, testimonials, case study
Email 5 (day 8): Soft offer or CTA to next step For each email:
- Subject line (3 variations — curiosity / benefit / question)
- Preview text (under 85 chars)
- Full email body
- CTA (text + button copy)
⬡Bulk email campaign — newsletterEmail
Write a monthly newsletter email:
Brand: [name] Topic/theme this month: [describe] Key content pieces to feature: [list 2-3 articles, updates, or offers] Audience: [subscriber type] List size: [approximate] Produce:
- Subject line (5 variations — A/B test options)
- Preview text for best subject line
- Full email (800-1200 words)
- Structure: personal opener → main story → 2-3 content sections → CTA → sign-off
- Footer: unsubscribe reminder, social links placeholder Tone: like writing to a friend, not a corporation
Video Content — AI Production Stack
🎬
Runway ML
AI Video · Should
Text-to-video, image-to-video, video editing with AI. Remove backgrounds, generate slow-mo, create video from text prompts. Gen-3 Alpha is best-in-class for short social videos and ad creatives.
Natural-sounding text-to-speech for video voiceovers, explainer videos, social content. Clone your own voice for consistent brand audio. Claude writes the script, ElevenLabs delivers it.
Free. Auto-captions, background removal, AI clip cutting, trending templates for Reels/TikTok. The fastest way to produce social video content. Desktop and mobile app. Claude writes the scripts.
Edit video by editing the transcript — delete words from the text and the video edits itself. AI filler word removal, eye contact correction, studio sound enhancement. Best for talking-head videos.
Write a 60-second video script for:
Platform: [Instagram Reel / TikTok / YouTube Shorts / LinkedIn] Topic: [topic] Goal: [educate / entertain / sell / build trust] Brand voice: [tone] CTA: [follow / visit link / comment / DM] Format:
- Hook (0-3 sec): visual direction + opening line — must stop the scroll
- Problem/context (3-15 sec): set up the tension
- Value delivery (15-50 sec): the main content in punchy sentences
- CTA (50-60 sec): clear ask Also provide:
- On-screen text suggestions for each section (under 6 words each)
- Thumbnail text suggestion
- Caption (150 words + 5 hashtags)
⬡Explainer video script (2 minutes)Video
Write a 2-minute explainer video script for:
Business/Product: [describe] Target audience: [who is watching] Problem it solves: [pain point] Key differentiators: [what makes it unique] Structure: Problem → Solution → How it works (3 steps) → Social proof → CTA
Spoken word: short sentences, natural rhythm, conversational
Indicate: [PAUSE], [VISUAL: describe what should show on screen], [GRAPHIC: text overlay]
End with: voiceover CTA + screen URL/contact
Total word count: ~280 words (fits 2 minutes at natural speaking pace)
Automations & n8n Workflows
These are the workflows that will save you the most time as a one-person marketing department. Build them in priority order.
New blog post published → Claude repurposes into 5 social formats → auto-queues in Buffer for the next 2 weeks
n8n + WordPress webhook + Claude + Buffer
Build custom
LATER
Brand Mention Monitor
Google Alerts → n8n → Claude sentiment analysis → categorises as positive/neutral/negative → Slack alert for negatives requiring response
n8n + Google Alerts + Claude + Slack
Build custom
LATER
Lead Nurture Sequence
New lead in CRM → Claude personalises email sequence based on lead source and industry → sends via Resend over 7 days
n8n + CRM + Claude + Resend
Build custom
LATER
Competitor Intelligence Report
Weekly: Firecrawl scrapes competitor blog + social → Claude identifies new content topics + ad angles → emails digest to you
n8n + Firecrawl + Claude + Gmail
Build custom
n8n Setup for Digital Marketing
⬡Build this workflow with ClaudeAutomation
I want to build an n8n automation workflow. Help me plan it. Workflow goal: [describe what should happen start to finish] Trigger: [schedule / webhook / new email / new review / form submission] Data sources: [what data it pulls — GA4 / Meta Ads / Sheets / etc] AI processing: [what Claude should do with the data] Output: [where results go — email / Slack / Sheets / Buffer / CRM] My n8n experience: beginner / I can follow step-by-step instructions Produce:
1. Workflow map (nodes in sequence, what each does)
2. Step-by-step build instructions
3. The exact Claude prompt to use in the AI node
4. What credentials/API keys I need
5. How to test it safely before going live
Digital Marketing Sub-Agents
Each sub-agent is a separate Claude chat with a specialist role. Open them alongside your main session. They are your virtual marketing team.
Before writing any content, auditing a page, planning a campaign
See prompt below
Ad Copy Writer
Generate ad copy variants, analyse performance, A/B test ideas
Before launching any campaign, refreshing ad creative, testing new angles
See prompt below
Analytics Interpreter
Translate data into insight, flag anomalies, write client narratives
After pulling reports, before client presentations, monthly reviews
See prompt below
Brand Voice Guardian
Review content for brand consistency, tone, and messaging alignment
Before publishing any content — the final quality gate
See prompt below
Reputation Manager
Draft review responses, monitor sentiment, crisis communication
When a negative review appears, crisis situations, weekly review management
See prompt below
Proposal Writer
Draft proposals, pitch decks, client presentations, case studies
New client pitches, contract renewals, award submissions
See prompt below
⬡SEO Strategist sub-agentSub-agent
You are a senior SEO strategist with 10 years experience in digital marketing.
You specialise in: keyword research, content strategy, on-page SEO, local SEO, and technical audits.
You do NOT write the final content — you produce briefs and strategies for a content writer to execute. My context:
- I am a Digital Marketing Manager at [CAC/client name]
- Target market: [describe]
- Current rankings: [describe what you know]
- Primary goals: [e.g. increase organic traffic by 30% in 6 months] When I give you a topic or URL, produce:
1. Target keyword + 5 supporting keywords
2. Search intent analysis
3. Content brief with heading structure
4. On-page optimisation checklist
5. Internal linking recommendations Always explain your reasoning so I learn from every brief.
⬡Analytics Interpreter sub-agentSub-agent
You are a senior digital marketing analyst. You turn raw data into clear, actionable insights.
Your job: receive marketing data, identify the story it tells, and produce client-ready narratives. Rules:
- Never use jargon without explaining it
- Always lead with the most important finding
- Separate what happened from why it happened from what to do about it
- Flag data anomalies and hypothesise causes
- All numbers should be compared to a baseline (previous period, industry average, or target) When I paste data:
1. Give me the 3-sentence executive summary first
2. Then the full analysis
3. End with 3 prioritised actions Audience: my client is [describe — business owner / marketing director / etc]
⬡Brand Voice Guardian sub-agentSub-agent
You are a brand voice guardian for [BRAND NAME]. Brand voice rules:
- Tone: [describe — e.g. professional but warm, confident but not arrogant]
- We always: [list 3-5 brand communication rules]
- We never: [list banned phrases and approaches]
- Target reader: [describe the audience]
- Brand personality: [3 adjectives] When I share content with you:
1. Rate it for brand voice alignment (1-10)
2. Flag any phrases that feel off-brand
3. Suggest alternative phrasing for anything that needs changing
4. Check for: clarity, tone, reading level (aim for grade 8), CTA strength
5. Give the content a final score: Publish Ready / Needs Minor Edits / Needs Rewrite Be direct. I'd rather know about a problem before it's published.
Proposals & Client Presentations
🎯
Gamma.app
AI Presentations · Must
Generate complete presentation decks from a text prompt or outline in 30 seconds. Beautiful templates, auto-layout, and AI content generation. Export as PDF or present live. Perfect for proposals.
AI automatically adjusts slide layouts as you add content. No more fiddling with design. Great for data-heavy reports and client presentations. Branding applied across all slides automatically.
Professional proposal templates with e-signature, tracking (see when clients open + read), and CRM integration. Claude writes the content, Proposify formats and delivers it.
Write a professional digital marketing proposal for:
Client: [business name + industry] Services proposed: [list: SEO / Social Media / Ads / Email / etc] Monthly retainer: [$ amount] Contract length: [e.g. 6 months minimum] Client's main pain point: [what problem they want solved] Our agency/your name: [name] Proposal structure:
1. Executive summary (1 page — the decision-maker reads only this)
2. Situation analysis (what we see wrong + what the opportunity is)
3. Proposed strategy (high-level approach for each service)
4. Deliverables (exactly what they get each month — be specific)
5. Timeline (months 1-3 focus, months 4-6 focus)
6. Investment (pricing table, what's included, what's extra)
7. Why us (3 differentiators + relevant case study reference)
8. Next steps (clear CTA to sign) Tone: confident, client-focused, not agency-speak
⬡Campaign results presentation scriptProposal
Write a script for presenting monthly campaign results to a client:
Client: [name + industry] Meeting length: [30 / 45 / 60 minutes] Key results this month: [paste your data] Wins to highlight: [what went well] Challenges to address: [what underperformed] Next month plan: [what you're doing differently] Script sections:
1. Opening (30 sec — set the narrative frame for the month)
2. Headline results (2 min — the 3 numbers that matter most)
3. Deep dive by channel (10 min — what happened + why)
4. What we're optimising (5 min — own the issues, present solutions)
5. Next month plan (5 min — show you're already working on it)
6. Q&A prompts (3 questions clients typically ask — with your prepared answers) Include: presenter notes, slide content suggestions, data visualisation recommendations
StudyVisaHub Tutorials
15 build tutorials — from blank repo to million-dollar platform. Every module: plan → tools → prompts → automations → monetisation.
🛡 Pre-Deploy Check
🗺 Project Overview
🌐 Website
📊 Data Dashboard
📧 Newsletter
🤝 Agent Directory
🏫 Provider Directory
📚 Course Finder
🎓 Scholarship Finder
✍ AI Blog
📱 Social Automation
👤 Student Registry
📋 Doc Checklist
🔔 DHA Alert System
📈 Strategy Reports
🧮 Student Tools
🖼 Infographics
💰 Monetisation
🛡 Pre-Deploy Security & Quality Check — SKILL.md
This is a reusable SKILL.md file for Claude Code. Install it once — Claude auto-runs a full security, quality, and performance check every time you ask it to review code before deploying. Works on every project, not just StudyVisaHub.
Why This Matters
One security miss in production can expose student data, API keys, or your Supabase database to the public internet. This skill runs every check automatically so you never forget a step.
How to Install the Skill
Step
Action
Who
1
In Antigravity, navigate to: ~/.claude/skills/ — create the folder if it doesn't exist
Create file: SKILL.md inside that folder — paste the content below
You
4
Test: in any project, say "run pre-deploy check on this file" — Claude should load and run it
Both
5
Share with son: commit the ~/.claude/skills/ folder to a shared GitHub repo — he clones it
You
SKILL.md — Copy and save this file
---name: pre-deploy-check description: Use when reviewing, auditing, or checking code before committing
or deploying. Triggers on: "check this", "review before deploy", "final check",
"security check", "debug this", "pre-deploy", "is this safe to deploy" allowed-tools: Read, Grep, Bash ---# Pre-Deploy Security & Quality Check Run ALL checks in sequence. Report by severity. ## 1. SECURITY (Critical — block deploy if found) - Exposed API keys/tokens: sk-, Bearer, apiKey=, secret=, password=, token=
- .env values hardcoded in client-side code
- SQL injection risks in database queries
- XSS: innerHTML or dangerouslySetInnerHTML with user input
- Unvalidated inputs used in queries or DOM
- CORS set to * in production
- Protected routes not validated server-side
- Supabase: RLS disabled on tables with user data ## 2. CODE QUALITY (Warning) - console.log() statements remaining
- TODO, FIXME, HACK comments
- Hardcoded URLs/values that should be env vars
- Async functions missing try/catch error handling
- Functions with no null/undefined input validation ## 3. PERFORMANCE (Note) - Images without width/height
- Redundant API calls on page load
- Missing loading states for async operations ## 4. MOBILE & ACCESSIBILITY (Warning) - Fixed widths that break below 768px
- Images missing alt text
- Form inputs missing labels ## REPORT FORMAT — use exactly this: 🔴 CRITICAL (block deploy)
🟡 WARNING (fix before next deploy)
🔵 NOTE (log and monitor)
✅ DEPLOYMENT READINESS: [READY / NOT READY] Ask: "Do you want me to fix any of these now?"
Quick Commands
# In any Claude Code / Antigravity session: "Run pre-deploy check on all changed files"
"Security check before I push this"
"Final review — is this safe to deploy?"
"Debug and security scan: [paste file]" # Claude will automatically:1. Load the pre-deploy-check skill2. Scan every file you paste3. Return structured Critical / Warning / Note report4. Offer to fix findings immediately
Run a complete pre-deploy security and quality check on this code. Check for ALL of the following and report by severity: 🔴 CRITICAL (block deploy):
- Any API keys, tokens, or secrets visible in the code (sk-, Bearer, apiKey=, password=)
- User inputs used in database queries without sanitisation
- Protected data accessible without authentication
- Supabase tables without RLS enabled 🟡 WARNING (fix soon):
- console.log() statements remaining
- Hardcoded values that should be in .env
- Async functions without error handling (try/catch)
- Functions that fail on null/undefined input 🔵 NOTE (monitor):
- Performance issues (redundant API calls, no loading states)
- Mobile breakpoint issues below 768px
- Images missing alt text For each finding: file name + line number + issue + exact fix. End with: ✅ DEPLOYMENT READINESS: READY / NOT READY Code to review: [PASTE ALL CHANGED FILES]
🗺 StudyVisaHub — Project Master Plan
The Vision
Australia's most trusted intelligence platform for international students, education agents, and providers. Data-driven. AI-powered. Built to be the go-to resource for the $40B+ Australian international education market.
Australia's most trusted, data-driven resource for international students, education agents, and providers. Must feel authoritative, clean, and trustworthy — like a government site but with personality.
Pages to Build
Page
Purpose
Key sections
Priority
Home
First impression + tool access hub
Hero with data counter, latest DHA update banner, top tools grid, newsletter CTA
Build first
Visa Information Hub
SEO traffic driver — informational intent
Subclass 500 guide, visa conditions, GTE requirement, evidence levels explainer
Build first
For Students
Student tools gateway
Links to: visa probability, PR checker, course finder, scholarship finder, doc checklist
Phase 2
For Agents
B2B acquisition
Agent directory, strategy reports, DHA alerts, compliance tools
Build the home page for StudyVisaHub.com.au. Brand: StudyVisaHub — Australia's trusted international education intelligence platform
Audience: International students, education agents, RTO/HE providers
Tone: Professional, data-driven, trustworthy — like a government resource with clarity Page sections:
1. Navigation: logo + Study in AU / For Agents / For Providers / Tools / Blog / Get Alerts
2. Hero: "Australia's International Education Intelligence Platform" + subheadline + 2 CTAs (Free Tools / Get DHA Alerts) + live stat counter (students in AU: 846,321)
3. Latest DHA Update banner: amber alert bar with placeholder for dynamic update
4. Tools grid: 6 cards — Visa Probability, PR Points Check, Course Finder, Scholarship Finder, Agent Directory, Strategy Reports
5. Newsletter signup: "Get DHA policy updates delivered to your inbox"
6. Stats section: 3 data points from latest DHA data
7. Footer: links + data sources disclosure + GDPR/privacy Stack: HTML/CSS/JS only. No frameworks. Dark navy + teal colour scheme.
Font: use DM Sans from Google Fonts.
Style: Premium, data-driven — NOT generic. Think Bloomberg meets gov.au.
Mobile: fully responsive.
SEO: include H1, meta title, meta description, structured data (WebSite schema)
📊 Student Visa Data Intelligence Dashboard
The Flagship Product
This becomes your premium SaaS offering — the dashboard agents and providers pay AUD $99-299/month for. It visualises DHA data, shows trends, and generates AI-powered strategic insights.
Build a data intelligence dashboard for StudyVisaHub.com.au. Data source: Supabase tables — visa_grants (country, grants, refusals, rate, month, year, sector), enrolments (state, sector, country, count, month, year) Dashboard layout:
- Top row: 4 KPI cards — total students, visa grant rate %, top source country, fastest growing sector
- Row 2: Line chart (grant rate over 12 months, filter by sector), bar chart (top 10 countries by enrolments)
- Row 3: Heat map by state (horizontal bar chart), pie chart by education sector
- Row 4: AI Insight button — "Generate strategic insights from this data" → calls Claude API Filters: date range selector, education sector dropdown, source country dropdown
Auth: check Supabase session — if not logged in, show "Sign up for free access" Stack: HTML/CSS/JS + Chart.js + Supabase JS client. No React.
Style: dark theme (match StudyVisaHub brand), data-dense but readable.
Performance: load data once on page load, cache in JS variables.
📧 Newsletter for Agents & Providers
Strategy
Two separate lists: "StudyVisaHub Weekly" for students (free, high volume) and "Agent Intelligence Brief" for agents/providers (paid premium, high value). The agent list becomes a revenue channel at AUD $49-99/month.
Newsletter Architecture
Newsletter
Audience
Frequency
Content
Platform
Model
StudyVisaHub Weekly
Students
Weekly
Visa tips, course highlights, policy updates, scholarship alerts
Beehiiv (free tier)
Free — audience building
Agent Intelligence Brief
Agents + providers
Monthly
DHA data analysis, evidence level updates, market intelligence, strategy tips
Beehiiv (paid)
AUD $49/mo subscription
Build Steps
Step
Task
Tool
Who
1
Create Beehiiv account + 2 publications
Beehiiv
You
2
Add signup forms to site — connect to Beehiiv via API
Beehiiv API + Supabase
Claude
3
Build 5-email welcome sequence for each list
Claude (content) + Beehiiv
Both
4
n8n automation: weekly email generation from DHA data → Claude drafts → you approve → sends
n8n + Claude + Beehiiv API
Both
5
Set up Stripe payment for agent newsletter subscription
Stripe + Beehiiv premium
You
⬡First agent newsletter editionNewsletter
Write the first edition of the "Agent Intelligence Brief" newsletter for StudyVisaHub. Audience: Australian education agents and registered training providers (RTOs)
Tone: Professional, data-driven, practical — we are their most reliable source of truth Structure:
1. Subject: "DHA Update: [most important change this month] — What agents need to know"
2. Opening: 2 sentences on why this month's data matters
3. Data spotlight: key number from latest DHA release with plain-English interpretation
4. Policy update: any recent evidence level changes or DHA announcements
5. Market intelligence: which source countries are trending up or down
6. Agent action item: one specific thing agents should do this week based on the data
7. Course spotlight: one course type showing strong visa grant rates
8. Closing CTA: upgrade to full dashboard access Use these real data points:
- 846,321 international students in AU (YTD Dec 2025)
- Visa fee: AUD $2,000 (world's highest)
- VET sector down 45% vs 2019
- HE sector up 14% vs 2019
- Evidence level update: effective 30 Sep 2025
🤝 Education Agent Directory Listing
Revenue Model
Free basic listing → AUD $29/month for featured → AUD $99/month for verified + leads. Agents pay to be visible to students searching for agents by location and specialty.
Directory Features
🔍 Search & Filter
Filter by: state, specialty (HE/VET/ELICOS), source country expertise, MARA registered, languages spoken
⭐ Agent Profile
Business name, MARA number, specialties, countries, languages, contact form, reviews, response time
📊 Performance Badge
Premium: "High Grant Rate Agent" badge based on anonymised DHA data integration
📝 Agent Submission
Self-service listing form. Manual verification of MARA number via DHA register before approval.
Build an education agent directory for StudyVisaHub.com.au. Data: Supabase table 'agents' with fields: id, business_name, mara_number, state, city, specialties (array), source_countries (array), languages (array), description, logo_url, tier, verified, rating_avg Features:
- Search bar: searches business_name and description
- Filters: state dropdown, specialty multiselect, verified toggle, language dropdown
- Agent cards: logo, name, state, specialties pills, languages, star rating, "Verified MARA" badge if verified=true, "Featured" gold border if tier=featured
- Click → agent profile page (route: /agents/[id])
- Agent profile: full details + contact form (submits to Supabase contact_requests table)
- Pagination: 20 per page, load more button
- Submission form: "List your agency" → form → sends to Supabase pending_agents table for manual review Featured agents appear first. Free agents sorted by state then name.
Stack: HTML/CSS/JS + Supabase. Mobile responsive.
🏫 Provider Listing Directory
Data Advantage
CRICOS data is public and free. Pre-populate the directory with all registered providers before anyone pays — this gives you instant credibility and search engine presence.
Build Steps
Step
Task
Tool
Who
1
Download CRICOS provider CSV (all registered AU providers)
Each course generates its own SEO page: "[Course Name] in Australia for international students". With 10,000+ CRICOS courses, this becomes a massive organic traffic driver.
Course Finder Features
🔍 Smart Search
Search by: course name, CRICOS code, field of study, state, provider type (HE/VET/ELICOS), duration, fees
🎯 PR Pathway Filter
Filter courses by: leads to MLTSSL occupation, eligible for 485 visa, regional study bonus points available
📊 Visa Intelligence
Show visa grant rate for students from [country] studying this course. Links to DHA data source.
💰 Fee Comparison
Compare annual tuition fees across providers offering the same qualification. Powered by CRICOS data.
⬡Build the course finderCourses
Build an Australian course finder for StudyVisaHub.com.au. Data: Supabase table 'courses' with fields:
id, course_name, cricos_code, provider_id, field_of_study,
qualification_type (Bachelor/Masters/Diploma/Certificate IV/etc),
duration_weeks, annual_fee_aud, state, delivery_mode,
pr_pathway (bool), mltssl_linked_occupation, regional_eligible (bool) Features:
1. Search: free text across course_name, field_of_study, provider name
2. Filters: qualification_type, state, field_of_study, pr_pathway toggle, max_fee slider
3. Results: card showing course name, provider, qualification, state, duration, annual fee, "PR Pathway" green badge if applicable
4. Course detail page (/courses/[id]): full details + provider info + PR pathway guide + "Apply" button (lead capture form)
5. SEO: each course has unique page with meta title: "[Course] at [Provider] for International Students — StudyVisaHub" Show 20 results per page. Sort by: relevance (default) / fee low-high / duration short-long
Stack: HTML/CSS/JS + Supabase. Mobile first.
🎓 Australian Scholarship Finder
High Intent Traffic
"Scholarship for international students in Australia" has very high search volume and strong conversion intent. This page drives newsletter signups and agent referrals.
Scholarship Data Sources
Australia Awards
Australian Government scholarship program for international students. Official data published annually.
# n8n workflow — runs monthly Trigger: Schedule (1st of each month)
→ Firecrawl: scrape scholarship pages from 20 top AU universities
→ Claude: extract structured data (name, amount, eligibility, deadline, URL)
→ Supabase: upsert into scholarships table
→ Resend: email digest to newsletter subscribers — "New scholarships found this month"
✍ Automated AI Blog Content Creator
The SEO Engine
Target 3 articles per week. Every DHA update is a content opportunity. Every visa subclass, occupation, and source country is a keyword cluster. 150 articles = significant organic traffic within 6 months.
Content Strategy
Content type
Volume
SEO target
Update trigger
DHA policy updates
Every update
"[policy change] Australia international students"
DHA alert system → auto-generate
Visa subclass guides
1 per subclass (20+)
"Student visa subclass 500 Australia"
Write once, update quarterly
Country-specific visa guides
1 per top 20 countries
"Student visa Australia from India"
Update when data changes
Course + career guides
1 per field of study
"Best courses in Australia for PR pathway"
Update when occupation lists change
PR pathway guides
1 per occupation cluster
"How to get PR from nursing course Australia"
Update when occupation lists change
Automated Blog Pipeline
# n8n workflow — DHA update → auto-generate blog post Trigger: DHA alert webhook (new policy detected)
→ Claude: "Write 1200-word SEO blog post about [update]. Target keyword: [keyword]. Include: what changed, who is affected, what agents should tell students, what providers should do."
→ Claude: generate meta title + description + 3 social post variants
→ Supabase: save as draft with status='pending_review'
→ Email you: "New draft ready for review: [title]" with approve/edit link
→ On approval: publish to site + auto-post social variants via Buffer API
⬡SEO blog post — DHA updateBlog
Write an SEO-optimised blog post for StudyVisaHub.com.au about this DHA update:
[DESCRIBE THE DHA UPDATE OR PASTE THE ANNOUNCEMENT] Target keyword: [primary keyword] Secondary keywords: DHA update, international students Australia, student visa 500, education agents Australia
Word count: 1,200-1,500 words
Audience: education agents and international students in Australia Structure:
- H1: Include primary keyword. Format: "[What changed]: What [audience] Need to Know"
- Introduction: Explain what changed in plain English (2 para)
- H2: What exactly changed (detail section)
- H2: Who is affected (students / agents / providers)
- H2: What this means for visa applications
- H2: What agents should tell their students right now
- H2: What providers need to update
- Conclusion: summary + link to DHA source + newsletter CTA
- Meta title: under 60 chars, include keyword
- Meta description: under 155 chars, include benefit and CTA Data citations: link to official DHA source for every claim.
Tone: authoritative but accessible. No jargon without explanation.
📱 Automated Social Media Content
Platform Strategy
Platform
Audience
Content type
Frequency
Automation level
LinkedIn
Agents, providers, industry
Data insights, policy updates, infographics
3x/week
80% automated
Instagram
Students, prospective students
Visa tips carousels, country spotlights, student stories
Daily
70% automated
Facebook
Students + parents
News, guides, community posts
5x/week
90% automated
TikTok
Young prospective students
60-sec visa tip videos, PR journey stories
3x/week
Script automated, edit manual
Social Automation Pipeline
# n8n workflow — daily social content generation Trigger: Schedule (6am daily)
→ Exa Search MCP: find trending Australia international education news
→ Claude: generate platform-specific posts for each channel:
- LinkedIn: data insight post (150 words + infographic prompt)
- Instagram: carousel script (6 slides)
- Facebook: news post with student angle
→ Canva API: generate visual from carousel script template
→ Buffer API: schedule posts at optimal times
→ Supabase: log content calendar entry # Special trigger: DHA update detected → Skip schedule → immediate post across all channels
→ Format: breaking news style — "⚠ DHA UPDATE: [what changed]"
👤 Student Registry — Lead Matching System
The Revenue Engine
Each registered student is a qualified lead. When matched to an agent or provider who converts them to an enrolment, you earn a referral fee. This is the core B2B revenue model at scale.
Student Registry Features
📝 Student Profile
Nationality, English level, budget, preferred study level, preferred state, desired PR outcome, timeline
🎯 Course Match
AI matches student profile to top 5 recommended courses based on: PR pathway, visa grant rate, budget, location preference
🎓 Scholarship Match
Automatically surfaces matching scholarships based on nationality, study level, and field of study
🗺 PR Pathway Planner
Shows: current points estimate → target occupation → recommended course → PR timeline → probability estimate
🤝 Agent Matching
Matches student to 3 agents specialising in their country and study level. Student contacts directly.
📊 Lead Dashboard
Agent/provider dashboard: view matched student profiles (anonymised until agent pays for contact details)
⬡Build student registry and matching systemRegistry
Build a student registry and matching system for StudyVisaHub.com.au. Student registration form collects:
- Personal: first name, email, country of origin, age
- Academic: current qualification, IELTS/PTE score (or estimate), field of interest
- Goals: preferred study level, preferred Australian state, budget range (AUD/year), PR goal (yes/no), timeline (next intake/6mo/12mo+)
- Source: how they heard about us On submission:
1. Save to Supabase students table
2. Trigger Claude API call: "Based on this student profile, recommend: top 3 courses, top 3 providers, whether they qualify for PR pathway, estimated points if PR goal, top 2 scholarships"
3. Send welcome email via Resend with personalised recommendations
4. Flag student as "new lead" in agent dashboard Agent lead dashboard (auth required, paid tier):
- See student profiles anonymised (country, study level, budget, PR goal)
- "Unlock contact details" button → Stripe payment of AUD $25-50 per lead
- Filter leads by country, study level, budget, PR goal Supabase RLS: students can only see their own profile. Agents can see anonymised profiles only until paid unlock.
📋 Document Checklist — Quality & Integrity Tool
High Value Free Tool
The document checklist is one of the highest-converting free tools — it generates newsletter signups and positions StudyVisaHub as essential for agents. The personalised version (by country + provider risk level) is a paid feature.
Checklist System Design
🌍 Country-based checklist
DHA requires different evidence levels by country. Checklist adapts based on student's country of origin and provider evidence level (high/medium/low).
📊 Evidence level integration
Links to DHA's evidence level data — shows how thorough the documentation needs to be for each country/provider combination.
✅ Interactive checklist
Students tick off documents as they prepare. Progress saved in Supabase. Shareable with their agent. Print/PDF option.
⚠ Integrity flags
Highlights common reasons for visa refusal: false documents, GTE failure, financial evidence issues. Education-first approach.
⬡Build the document checklist toolChecklist
Build an interactive document checklist tool for StudyVisaHub.com.au. Step 1: User inputs country of origin (dropdown, 20 most common) + provider risk level (high/medium/low, from dropdown)
Step 2: System generates personalised checklist Checklist categories always shown:
1. Identity documents (passport, birth certificate)
2. Academic qualifications (certified transcripts, IELTS/PTE results)
3. Financial evidence (bank statements, sponsor letter — amount varies by country/duration)
4. GTE (Genuine Temporary Entrant) statement guide
5. Health insurance (OSHC requirement)
6. Enrolment confirmation from provider (CoE) Country-specific additions (show for high-risk countries):
- Additional financial evidence requirements
- Extra identity verification
- Sponsor statutory declaration Interactive features:
- Each item has a checkbox (saves to localStorage)
- Info icon explains why each document is required
- Progress bar: "X of Y documents ready"
- Print checklist button (PDF)
- Email checklist button (saves email to Supabase + sends via Resend)
- "Get personalised advice" CTA → agent matching Important note at top: "This checklist is a guide only. DHA requirements change. Always verify with a registered migration agent or directly with DHA."
🔔 DHA Alert System — Automated Policy Monitor
The Most Valuable Feature
Agents pay thousands per year for consultants who tell them about DHA updates. You give this away free (basic) and charge for premium. This is the single fastest path to first paying subscribers.
# n8n workflow — daily DHA monitoring Trigger: Schedule (7am daily) Branch 1: DHA page change detection
→ Firecrawl: scrape homeaffairs.gov.au/news-media/current-alerts-and-advisories
→ Compare with yesterday's version (stored in Supabase)
→ If changed: extract new content
→ Claude: "Summarise this DHA update in plain English for education agents. Include: what changed, who is affected, what agents must do, timeline."
→ Supabase: store alert with severity (critical/important/info)
→ Email you immediately with full analysis
→ Send to premium subscribers within 2 hours
→ Queue blog post draft for your review Branch 2: Monthly data check (runs 1st of month)
→ Check education.gov.au for new monthly summary PDF
→ If new: download + extract key metrics
→ Claude: generate data analysis + newsletter draft
→ Email you for review Branch 3: Weekly data.gov.au check
→ Check for new CSV files in student visa dataset
→ If new: download → update Supabase → trigger dashboard refresh # Email alert format Subject: ⚠ DHA UPDATE [CRITICAL/IMPORTANT]: [what changed]
Body: Plain English summary + affected parties + required actions + source links
Alert Tiers — Monetisation
Tier
Price
What they get
Free
$0
Weekly digest email, 48hr delay on alerts
Agent Essential
AUD $49/month
Real-time alerts, full analysis, monthly data report
Provider Intelligence
AUD $149/month
Above + evidence level tracking for their CRICOS code, student demand intel
⬡Build the DHA alert n8n workflowAlerts
Help me build an n8n automation workflow to monitor DHA website for policy updates. Workflow goal: Monitor 5 Australian government education URLs daily, detect any content changes, generate an AI analysis report, and email it to me + subscribers URLs to monitor:
1. https://www.homeaffairs.gov.au/news-media/current-alerts-and-advisories
2. https://www.education.gov.au/international-education-data-and-research
3. https://data.gov.au/data/dataset/student-visas Detection method: Firecrawl MCP to scrape each URL, compare hash/content with previous day (stored in Supabase table 'page_snapshots') On change detected:
1. Extract the new content
2. Send to Claude API: "You are an expert in Australian international education policy. Analyse this DHA update and produce: (a) 2-sentence plain English summary, (b) who is affected, (c) what agents must do immediately, (d) what providers must update, (e) urgency level: CRITICAL/IMPORTANT/INFO"
3. Store in Supabase 'alerts' table with: url, detected_at, summary, full_analysis, urgency
4. Send email via Resend to my address AND to all subscribers in Supabase 'alert_subscribers' where tier='premium' Tools needed: n8n Cloud, Firecrawl API, Claude API, Supabase, Resend
Give me: workflow node map + step-by-step build guide + Claude prompt for the AI analysis node
📈 Strategy Reports for Agents & Providers
High-Value Product
A data-driven strategy report telling an agent or provider which courses, countries, and regions to focus on is worth AUD $299-999 to them. It takes you 30 minutes with AI. At 10 reports/month = AUD $3,000-10,000/month.
Report Types
Report
Who buys it
Price
What it contains
Agent Market Intelligence Report
Education agents
AUD $299
Top source countries by grant rate, best courses for their specialty, visa risk flags, monthly trend analysis
Provider Enrolment Strategy Report
RTOs + universities
AUD $499
Which source countries to target, grant rate by their CRICOS code, comparable provider benchmarking, agent recruitment strategy
Annual Market Outlook
Both
AUD $999
Full-year DHA data analysis, sector trends, country-by-country breakdown, 12-month forward forecast
⬡Generate agent strategy reportReports
Generate a data-driven strategy report for an education agent. Agent profile:
- Location: [state] - Specialisation: [VET / HE / ELICOS / mixed] - Main source countries: [list 3-5] - Current volume: [approx students/year] - Goal: [grow volume / improve grant rate / enter new market] Available data (paste or describe): [paste DHA data: grant rates by country, sector enrolment trends] Report structure:
1. Executive summary (which 3 markets have the best opportunity right now)
2. Source country analysis — for each of their main countries: grant rate trend, risk level, recommended course types
3. Course opportunity analysis — which qualifications are growing vs declining
4. Visa risk mitigation — evidence level status, documentation recommendations
5. New market opportunity — 2 source countries they're not working with that show strong indicators
6. 90-day action plan (3 specific actions with expected impact)
7. Data sources + disclaimer Format: professional PDF-ready. Use data tables, bullet points, clear headings. Include a disclaimer that this report uses publicly available data and is not migration advice.
🧮 Student Tools — Visa Probability, PR Checker, Strategy Planner
🎲
Visa Probability Checker
Students input profile → get estimated approval likelihood
How it works
Student inputs: nationality, study level, IELTS score, bank balance, course duration, provider evidence level. Claude calculates weighted probability based on DHA grant rate data for that country/sector combination. Shows: percentage estimate, main risk factors, how to improve chances.
⬡Build visa probability tool
Build a visa probability checker for StudyVisaHub.com.au. Form inputs:
- Nationality (dropdown — top 30 source countries)
- Study level (HE / VET / ELICOS / Schools)
- Course duration (dropdown: 6mo / 1yr / 2yr / 3yr / 4yr+)
- IELTS/PTE score (dropdown ranges)
- Bank balance available (dropdown ranges in AUD)
- Previous visa refusals (yes/no)
- Agent assisting (yes/no) On submit: POST to /api/visa-check (Cloudflare Worker) which:
1. Looks up grant rate for that nationality + sector from Supabase visa_grants table
2. Applies modifiers: +/- based on IELTS, finance, prior refusals
3. Calls Claude API: "Based on this student profile and a base grant rate of [X]%, estimate approval probability and top 3 risk factors to address. Be honest but constructive."
4. Returns: probability estimate (%), risk factors (3), improvement tips (3), recommended action Display: probability meter (visual), traffic light (green/amber/red), risk factor cards, "Speak to an agent" CTA Important disclaimer: "This is an estimate only based on historical data. DHA assesses each application individually. This is not migration advice."
🏅
PR Points Calculator
Interactive points test for Subclass 189/190/491
The official points test is on DHA's website but it's not user-friendly. Build a better UX version that shows exactly how to earn more points and which visa pathway is best for the student's current score. Minimum 65 points required; competitive invitations at 75-90 points (2025-26 data).
The highest-value student tool. Student inputs current situation → receives a complete 2-5 year roadmap: which course to enrol in, which state to study in (regional bonus?), which occupation to target, how many points they'll accumulate, when to apply for PR, estimated timeline to citizenship.
⬡Generate student PR strategy
Generate a personalised PR pathway strategy for this student: Student profile:
- Nationality: [country] - Age: [age] - Current qualification: [highest degree] - English level (IELTS equivalent): [score] - Work experience in skilled occupation: [years] - Desired field of study in Australia: [field] - PR goal: Yes
- Budget: AUD [budget]/year
- Partner: [single / partner with skills / partner without skills] Using the 2025-26 Australian points test:
1. Calculate current estimated points score
2. Identify the best visa pathway (189/190/491)
3. Recommend the optimal course in Australia (maximise points + employment)
4. Recommend state (regional = +15 points on 491?)
5. Project points score after completing recommended course
6. Show 5-year timeline: Arrive → Study → Graduate → Work → Apply PR → Grant
7. Flag any risks (age clock, occupation list changes, visa fee costs)
8. Estimate total cost of PR pathway (tuition + visa fees + living) Important: End with "This is an educational estimate only. Consult a registered migration agent (MARA) before making any decisions." Data reference:
- Points test: https://www.anzscosearch.com/points-test/
- Occupation list: https://immi.homeaffairs.gov.au/visas/working-in-australia/skill-occupation-list
🖼 Infographics — Job & Career Pathway for PR
Virality Strategy
Infographics are the most shareable content in the international education space. One well-designed infographic shared by agents in their WhatsApp groups can drive thousands of site visits. Target: agents as distribution channel.
Infographic Series to Create
Infographic
Target audience
Distribution
Tools
Australia PR Pathway: Course → Job → PR
Students + agents
LinkedIn, Instagram, WhatsApp (agent groups)
Canva AI / Claude + Canva
Top 10 PR-Pathway Courses in Australia 2025
Students
Instagram carousel, Pinterest, blog
Canva AI
Australian Visa Grant Rate by Country (2025)
Agents + providers
LinkedIn, newsletter, downloadable PDF
Chart.js → Canva export
Australia PR Points Calculator Visual
Students
All social platforms
Canva AI template
Study → Work → PR: 5-Year Timeline
Prospective students
High-traffic blog post, social
Canva AI
Evidence Level Guide: Am I High/Medium/Low Risk?
Agents
LinkedIn, newsletter
Canva + Claude data
Production Workflow
# Infographic production pipeline 1. Claude: create infographic content brief (what data, what story, what layout)
2. Canva AI: "Create infographic titled [X] with these data points: [Y]"
OR Claude Code: generate SVG infographic directly
3. Brand: apply StudyVisaHub colour palette (navy + teal + white)
4. Export: 4 formats — square (Instagram), tall (Pinterest/Stories), wide (LinkedIn), A4 (PDF download)
5. Publish: upload to site as downloadable lead magnet → email gate → collect leads
6. Distribute: Buffer schedule across all platforms
⬡PR Pathway infographic content briefInfographic
Create an infographic content brief for "Australia PR Pathway: Study → Work → Permanent Residency" Target audience: International students considering Australia
Goal: Show the complete journey from student visa to PR to citizenship in one visual Sections of the infographic:
1. Header: "Your Australia PR Journey — From Student to Permanent Resident"
2. Timeline visual: 5 stages with estimated durations
Stage 1: Arrive on Student Visa (Subclass 500) — 1-4 years study
Stage 2: Graduate — apply for Temporary Graduate Visa (Subclass 485) — 2-4 years
Stage 3: Build skills + work experience in nominated occupation
Stage 4: Submit Expression of Interest (EOI) — Points Test — 65+ points needed
Stage 5: Receive Invitation → Lodge PR Application → Grant
3. Points summary box: key categories and max points (use real 2025-26 data)
4. Top 5 courses for PR pathway (include occupation examples)
5. Disclaimer: "This is a simplified guide. Consult a MARA registered agent."
6. Footer: StudyVisaHub.com.au branding Data to include:
- Minimum 65 points for pool entry (competitive rounds 75-90 in 2025)
- 189 (independent), 190 (state nominated), 491 (regional) pathway options
- 485 visa: 2-4 years depending on study level and region
- Total pathway: typically 5-8 years from student visa to citizenship Produce: headline, all text for each section, specific numbers to include, visual layout description (so I can brief Canva AI or a designer)
💰 Monetisation Blueprint — The Million Dollar Platform
The Vision
StudyVisaHub serves 3 paying audiences: students (volume), agents (premium), and providers (enterprise). Each layer adds a revenue stream. At scale, this is a AUD $1M+ business.
Revenue Streams — All 12
Revenue stream
Model
Price point
Scale potential
Phase
Agent Intelligence Newsletter
Subscription
AUD $49/mo
1,000 agents = $49K/mo
Phase 1
DHA Alert Premium
Subscription
AUD $49-149/mo
500 subscribers = $50K/mo
Phase 1
Agent Directory Featured Listing
Listing fee
AUD $29-99/mo per agent
500 agents = $25K/mo
Phase 2
Provider Directory Enhanced Listing
Listing fee
AUD $199-499/mo
200 providers = $60K/mo
Phase 2
Data Intelligence Dashboard
SaaS
AUD $99-299/mo
500 subscribers = $100K/mo
Phase 3
Strategy Reports (one-off)
Report sales
AUD $299-999 each
100 reports/mo = $50K/mo
Phase 3
Student Lead Referrals
Lead fee
AUD $25-100 per lead
500 leads/mo = $25K/mo
Phase 3
Course/provider referral commissions
Affiliate %
5-15% of first-year tuition
100 enrolments × $2K = $200K/yr
Phase 4
Infographic + data downloads
Lead magnet → upsell
Free → upgrade
Newsletter + dashboard conversions
Phase 2
Display advertising
CPM/CPC
AUD $5-20 CPM
100K pageviews/mo = $2-5K/mo
Phase 3
White-label reports for agents
Agency tier
AUD $499/mo
50 agencies = $25K/mo
Phase 4
Student visa consultation referrals
Partner fee
AUD $50-200 per booked consultation
200/mo = $20K/mo
Phase 4
Path to $1M ARR
Phase
Milestone
MRR target
How to get there
Month 1-3
Launch + first 100 subscribers
AUD $5,000/mo
DHA alerts free tier drives signups → convert 10% to paid $49/mo
Month 3-6
Directories live + 50 paying agents
AUD $15,000/mo
Agent directory featured listings + newsletter upgrade path
Find 20 active Facebook groups and WhatsApp networks where Australian education agents gather. Be the person who shares the best DHA data summaries. Every post links back to studyvisahub.com.au.
🔗 MARA Agent Outreach
DHA publishes the full registered migration agent list. Email every MARA agent in Australia announcing your free alert service. Even 1% converting to paid = hundreds of subscribers.
📚 CRICOS Provider Cold Email
Email every Australian RTO and university international office. Subject: "Free DHA grant rate data for [their CRICOS code]". Hook with free data → sell dashboard.
🎤 Industry Conferences
ISANA, AIRC, PIER — present the platform at Australian international education conferences. One conference = hundreds of qualified agent and provider contacts.
📊 LinkedIn Thought Leadership
Post one DHA data insight per week on LinkedIn. Agents and providers follow. Every post ends with "Full analysis at StudyVisaHub.com.au". Build audience before monetising.
🤝 Migration Agent Partnerships
Partner with 5-10 MARA agents who will recommend the platform to their clients. Revenue share: you send them student leads, they send you subscribers.
The $1M Insight
The platform that wins is not the one with the most features — it's the one agents trust most. Every DHA update you send faster and more accurately than anyone else builds that trust. Start with the alert system. Let everything else grow from the trust you build there.
My Learning Path
Biv's personalised roadmap from Advanced Power User → AI Solutions Architect & Automation Specialist. Skills gap tracker, video resources, and milestone map — all in one place.
🎯 My Job Title
📊 Skills Gap Map
🗺 Full Roadmap
💻 Technical Skills
⚙ Automation Skills
📣 Marketing AI
▶ All Courses
✅ Progress Tracker
Your Job Title — The Honest Answer
Your Title
AI Solutions Architect & Automation Specialist — You design and build AI-powered product systems, combining Claude, n8n, Supabase, and automation pipelines to solve real business problems. You sit at the intersection of product thinking and AI execution — where most developers can't think, and most marketers can't build.
Why This Title Fits You Better Than the Others
Title
What it really means
Your fit today
After StudyVisaHub
AI Automation Engineer
Writes, maintains, and debugs code independently. Owns the technical stack.
~35% — you direct, Claude writes
~70% — gaps closing fast
AI Automation Specialist
Designs workflows, selects tools, manages AI systems. Less coding than engineer.
~65% — strong in design, weak in execution
~85% — very credible
AI Solutions Architect
Defines the problem, architects the solution, orchestrates the build. Strategic + technical.
~75% — this is your real strength
~92% — fully credible
AI Solutions Builder
Hands-on product creator. Concept → deployed tool using AI as the build engine.
~70% — StudyVisaHub proves this
~90% — proven track record
Your Elevator Pitch (copy this)
⬡Your professional bio — after StudyVisaHub is liveIdentity
I design and build AI-powered product systems — combining Claude AI, n8n automation, Supabase, and Vercel to turn complex problems into working digital products. I've shipped StudyVisaHub.com.au, Australia's intelligence platform for international education — featuring automated DHA policy alerts, a student visa data dashboard, course and scholarship finders, and AI-generated strategy reports for education agents and providers. My background is digital marketing (SEO, Meta Ads, Google Analytics, content strategy, brand management) which means I build systems that don't just work — they're designed to acquire users, retain them, and generate revenue. I work across the full stack: from product strategy to Supabase schema, from Claude Projects to n8n automation workflows, from UI design to SEO and paid acquisition. Job title: AI Solutions Architect & Automation Specialist
Location: Darwin, Australia
Available for: consulting, fractional CMO/CTO roles, platform builds
The 3 Gaps That Matter Most Right Now
Gap
Why it matters
How to close it
Time needed
Read JS well enough to spot bugs
You can't debug what you can't read. Claude produces JS — you need to verify it.
The DHA alert system is your first. Do it manually — not from a template.
n8n free tier + follow the SVH alert tutorial in this playbook
1 full day
Write one Supabase RLS policy yourself
Supabase without RLS = public database. You must understand this before launch.
Supabase RLS guide + apply it to the students table in SVH registry
2–3 hours
Skills Gap Map — Where You Stand Today
How to use this
Each skill has a checkbox. Tick it when you've practiced it in a real project — not just read about it. The StudyVisaHub build is designed to tick most of these naturally.
AI Tools & Workflow
☐
Skill
Your level
How to learn
Status
Multi-model prompting (Claude + GPT + Gemini)
Strong — already active
Keep practising daily across all 3
Strong
Claude Projects + Skills + CLAUDE.md (in practice)
Full Roadmap — Power User → AI Solutions Architect
The Strategy
You learn by building. The Anthropic courses give you the conceptual foundation. The StudyVisaHub build gives you the practical proof. Every milestone below maps to a real thing you ship.
Phase 1 — Foundation (Weeks 1–4)
Do the Anthropic courses. Set up your full Claude Code environment. Build your first n8n workflow. These are prerequisites — do not skip.
You don't need to write HTML from scratch — you need to read it. Goal: open any file Claude produces and understand what each section does. Do the first 30 exercises on freeCodeCamp — takes about 3 hours.
freeCodeCamp: Responsive Web Design cert (first 30 exercises) · MDN: HTML basics + CSS layout
You only need modules 1–4 of the freeCodeCamp cert: variables, functions, arrays, objects. That covers 95% of what Claude writes. After that, just ask Claude to explain every function it produces until you can read them yourself.
Modules: Basic JS · ES6 · Regular Expressions · Basic Data Structures — stop there
Every API returns JSON. The DHA data, Supabase queries, Claude API responses — all JSON. Takes 30 minutes to learn. Then practice: every time Claude makes an API call, look at the response and trace the data path yourself.
W3Schools JSON intro (30 mins) · Then: paste any API response into a new chat and ask Claude to explain the structure
Critical before SVH launches. RLS (Row Level Security) determines who can read/write which data. Without it, your database is publicly accessible. Read the RLS guide, then write at least one policy yourself — don't let Claude write it without understanding what it does.
Quickstart · Auth guide · RLS guide · Then: apply RLS to SVH students and agents tables manually
You only need 6 commands: git status, git add, git commit, git push, git checkout [hash] -- file, git log --oneline. The rest Claude handles. The "Oh Shit Git!" site is the best resource for when things go wrong.
Atlassian basics · Practice: run all 6 commands on SVH repo at least once each
Cloudflare Workers let you put a serverless function between your frontend and any API — hiding your Claude API key from the browser. Required for SVH's AI features. Claude writes the Worker code — you just need to understand what it does and how to deploy it.
Quickstart guide (30 mins) · Deploy one Worker for SVH's Claude API proxy · Verify key is not visible in browser DevTools
Required for SVH's paid tiers. The checkout is easy — Stripe handles it. The webhook (when payment succeeds, update user tier in Supabase) is the tricky part. Do the quickstart in test mode. Claude writes the code — you need to understand the payment → webhook → database update flow.
Checkout quickstart · Webhook guide · Test a payment in Stripe test mode before touching production
Automation Skills — n8n, Webhooks & Pipelines
The Priority
Your first n8n workflow should be the SVH DHA Alert System. It touches every automation concept you need: scheduled trigger, HTTP request, data comparison, Claude API call, conditional logic, email output. Build this one and you've learned 70% of n8n.
Start with the free n8n cloud tier — no install needed. Build the DHA alert workflow from the SVH tab. Then the weekly marketing report workflow. By the third workflow you'll be comfortable with any workflow.
n8n Fundamentals course (free on docs.n8n.io) · Then build: DHA monitor → Weekly GA4 report → Social scheduler
A webhook is a URL that receives data when something happens somewhere else. Your WordPress site publishes a blog post → sends data to n8n webhook → Claude repurposes it → Buffer schedules posts. This pattern powers most of your automation ideas.
Webhooks explained video (10 mins) · n8n Webhook node docs · Build: blog post → social repurpose pipeline using webhook trigger
n8n has a built-in AI Agent node that connects directly to Claude. You give it a prompt and a set of tools — it runs autonomously. This is the engine behind your DHA alert analysis, content generation, and reporting workflows.
n8n AI Agent node docs · Add to DHA alert workflow: HTTP request gets data → AI Agent node analyses it → email sends
Priority Automation Builds — In Order
☐
Workflow
What you learn
Template
1. DHA Alert Monitor (build from scratch)
Scheduled trigger, HTTP request, page comparison, Claude API, conditional email
Webhook trigger, Claude draft, email approval flow, conditional posting
Build custom — use SVH Reputation tab prompt
5. Blog → 5 social formats pipeline
WordPress webhook, Claude repurposing, Buffer API, content calendar automation
Build custom — use SVH Blog tab
Digital Marketing AI — Building on Your Existing Strengths
Your Advantage
You already have 10+ years of digital marketing expertise. The AI layer doesn't replace that — it multiplies it. Your job is to add AI workflows on top of skills you already have, not start from scratch.
First n8n workflow (DHA alert) live and sending emails
Not started
Phase 2
Supabase RLS written and understood
Not started
Phase 3
SVH data dashboard + student registry live
Not started
Phase 3
First paying subscriber (AUD $49/mo)
Not started
Phase 4
First strategy report sold (AUD $299+)
Not started
Phase 4
LinkedIn updated — first enquiry from SVH case study
Not started
When you can say this
"I designed and built StudyVisaHub.com.au — Australia's intelligence platform for international students and education agents. It runs automated DHA policy alerts, a live data dashboard, AI-generated strategy reports, and student-to-agent matching. I built it using Claude Code, n8n, Supabase, Vercel, and the Anthropic API. It has paying subscribers."
That is your job title proven. AI Solutions Architect & Automation Specialist.
10-Day Sprint
Launch StudyVisaHub.com.au in 10 days. Theory + practical paired daily. Every task is specific, timed, and sequenced. Start tomorrow.
⚙ Day 0: Setup
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
Day 7
Day 8
Day 9
Day 10 🚀
📁 Obsidian Setup
🌐 Deploy to Mediamantra
🔍 Course Gaps Fixed
⏱ Daily Rhythm
🆘 When Stuck
Day 0 — Setup Everything Before You Start
Do this today
Day 0 is not learning — it is setup. Every account, folder, and tool must be ready before Day 1. A bad setup means you lose 2 hours on Day 1 troubleshooting instead of building.
# Your Obsidian vault structure for this project# Mirror this in Google Drive so it syncs automatically 📁 ObsidianVault/
└── 📁 StudyVisaHub-Sprint/
├── 📄 00-PROJECT-OVERVIEW.md # Paste CLAUDE.md here too ├── 📄 01-DAILY-JOURNAL.md # One entry per day ├── 📄 02-LEARNINGS.md # memory.md — paste end-of-session outputs here ├── 📄 03-PROMPTS-THAT-WORKED.md # Save every prompt that got great output ├── 📄 04-BUGS-AND-FIXES.md # What broke + how you fixed it ├── 📄 05-DECISIONS.md # Why you made key technical decisions ├── 📄 06-SPRINT-CHECKLIST.md # Copy of this sprint tab └── 📁 code-snippets/ # Save useful code Claude generated# Google Drive mirror path: Google Drive > AI-Builder > StudyVisaHub-Sprint > [same structure] # Antigravity/Claude Code project path: ~/projects/study-visa-hub/ (local)
→ synced to GitHub: github.com/[yourname]/study-visa-hub
→ deployed to: studyvisahub.com.au via Vercel
Claude Code Init — Run This on Day 0
# In Antigravity terminal: mkdir ~/projects/study-visa-hub
cd ~/projects/study-visa-hub
git init
git remote add origin https://github.com/[yourname]/study-visa-hub.git
claude init # Verify it worked: /status # should show project initialised# Create your CLAUDE.md immediately:# (Claude will help you fill it — paste the template from Phase 4 tab)
Daily Opening Prompt — Use Every Morning
⬡Daily standup — open every session with thisDaily
Good morning. I am Biv, building StudyVisaHub.com.au.
Sprint Day: [DAY NUMBER] of 10
Today's target: [paste today's target from sprint tab] Yesterday I completed: [what you actually finished]
Yesterday I got stuck on: [anything that wasn't resolved]
Current repo state: [working / something is broken] Today's session task: [ONE specific thing from today's plan] Rules for this session:
- Complete files only — no snippets
- Explain every change in plain English (I am a non-developer)
- Flag before touching anything outside today's task scope
- After each task: tell me what I learned and what to add to my journal CLAUDE.md content: [paste current CLAUDE.md]
Current file: [paste file if editing one]
Rapid Learning Protocol — Learn Any Concept in 20 Minutes
⬡20-minute concept learning promptLearning
Teach me [CONCEPT] in 20 minutes. I am a non-developer. Structure:
1. One-line plain English definition (no jargon)
2. One real-world analogy (not from software)
3. One concrete example from StudyVisaHub specifically
4. The ONE thing I must understand to use this safely
5. The ONE mistake beginners make with this
6. Show me what it looks like in code (30 lines max)
7. What should I add to my Obsidian learnings file? Keep it fast. I am time-constrained. No theory beyond what I need to build.
Day 1 — Foundation + Environment
Goal today
All accounts live. Repo created. Claude Code running. First CLAUDE.md written. First commit made. Claude 101 started. By end of day: SVH repo exists at github.com and Claude can read it.
This is the day you build your first real n8n workflow. The DHA alert system is the most valuable feature on SVH. It also teaches you 70% of what you need to know about automation in one build.
Variables, functions, arrays. These are in every API call Claude writes.
4:00pm
Build: Alerts page on SVH website
SVH Alerts tutorial
Page that shows the latest DHA alerts from Supabase alerts table. Claude builds it.
6:00pm
Test: trigger workflow manually, confirm email received
—
Run workflow manually. Check your email. Check Supabase for the alert record.
Day 5 — Directories (Agent + Provider)
Goal today
Agent directory and Provider directory live with CRICOS data pre-loaded. Search and filter working. Agent profile pages accessible. First listing submission form live.
⏱
☐
Task
Theory / Resource
Practical
9:00am
Watch: Claude Code in Action — Setup + Context lessons
Objects and basic data structures — this is how Supabase data is structured
5:00pm
Pre-deploy check + commit both directory pages
Pre-deploy skill
Run check. Action warnings. Commit: "add agent + provider directories — CRICOS data loaded"
Day 6 — Course Finder + Scholarship Finder + Student Tools
Goal today
Course finder live with CRICOS course data. Scholarship finder with first 20 scholarships. Visa Probability checker and PR Points Calculator built and working.
Deploy Worker. Verify Claude API key is NOT visible in browser DevTools network tab.
Day 7 — Data Dashboard + Student Registry
Goal today
Data dashboard live with real DHA data. Student registry form live. Basic lead matching working. This is the day SVH starts looking like a real product.
Focus on: keyword intent + on-page SEO. These apply to every SVH page.
5:00pm
Add SEO meta tags to all pages built so far
SVH Website tutorial → SEO section
Claude adds meta title, description, og tags to every page. Verify in browser <head>.
Day 8 — Blog + Social Automation + Document Checklist
Goal today
AI blog system with first 5 posts published. Social automation n8n workflow live. Document checklist tool complete. SVH is now a full content platform.
⏱
☐
Task
Theory / Resource
Practical
9:00am
Build: Blog page with Supabase CMS
SVH tab → AI Blog tutorial
Create posts table in Supabase. Blog list page. Individual post pages. Claude builds all 3.
10:30am
Generate + publish 5 SEO blog posts with Claude
Digital Marketing tab → Content → blog post prompt
Use the SEO blog post prompt 5 times. Topics: Student Visa 500 guide, PR pathway, evidence levels, VET vs HE, top source countries.
Import template. Configure for SVH topics. Connect to Buffer. Test with one manual trigger.
3:00pm
Build: Document Checklist tool
SVH tab → Doc Checklist prompt
Interactive checklist. Country dropdown. Saves progress in localStorage. Print/email option.
5:00pm
Security review: run Security Reviewer sub-agent
Prompts tab → Security Reviewer prompt
Paste all new files into the security reviewer. Action every Critical and High finding.
Day 9 — Payment Layer + Strategy Reports + Full Test
Goal today
Stripe checkout live for Agent newsletter tier. First strategy report template ready. Complete end-to-end test of every page. Pre-launch checklist completed.
Claude generates first Agent Market Intelligence Report from DHA data. Save as PDF. Add to site.
2:00pm
Full end-to-end test — every page, every form
Phase 7 → Pre-deploy checklist
Click every link. Submit every form. Check every console error. Test on mobile. Document bugs.
4:00pm
Fix all bugs found in testing
Prompts tab → Fix a bug prompt
Paste each bug to Claude with exact console error. Fix. Re-test. Commit each fix separately.
5:30pm
Final pre-deploy check on ALL files
Pre-deploy skill
"Run pre-deploy check on all changed files" — zero CRITICAL findings required before Day 10
Day 10 🚀 — Launch Day
Launch Day
StudyVisaHub.com.au goes live today. This is not a soft launch — you are announcing it. Every page works. The DHA alert system is running. The newsletter is live. You have a real product.
⏱
☐
Task
Detail
8:00am
Final security scan — all production files
Zero API keys in code. Zero console errors. RLS on all Supabase tables. .env not committed.
9:00am
Final mobile test on real phone
Open studyvisahub.com.au on your actual phone. Click every page. Test every form.
10:00am
Deploy to production — final Vercel push
git push → Vercel auto-deploys → confirm studyvisahub.com.au loads the final version
10:30am
Set up all n8n workflows on live URLs
Update n8n workflows to use production URLs not localhost. Test each one triggers correctly.
11:30am
Announce on LinkedIn
Use the Digital Marketing → Proposals tab → Campaign results prompt to write the announcement post. Include: what it does, who it's for, link, screenshot.
12:00pm
Email your agent network
Use the Digital Marketing → Email → newsletter prompt. Subject: "Free DHA alerts just launched — studyvisahub.com.au"
2:00pm
Social posts scheduled in Buffer
LinkedIn + Facebook + Instagram announcing launch. Use the 30-day content calendar prompt to plan next 30 days.
4:00pm
Update your job title on LinkedIn
AI Solutions Architect & Automation Specialist. Add SVH to your profile as a project.
5:00pm
Final sprint journal entry
Write in Obsidian: what you built, what you learned, what surprised you, what's next. This is your case study.
You just shipped
In 10 days you went from no website to a live intelligence platform with automated DHA alerts, a data dashboard, course finder, scholarship finder, student registry, agent directory, and a paying subscription product. That is the proof of the title: AI Solutions Architect & Automation Specialist.
Obsidian + Google Drive + Antigravity Setup
Your Local Knowledge System
Obsidian is your brain. Google Drive is your backup. Antigravity reads your project files. This setup means your learnings, notes, and code are all in one connected system.
In Obsidian: New Vault → choose a folder inside your Google Drive → name it "AI-Builder"
3
Create sprint folder structure
New folder in vault: StudyVisaHub-Sprint → create the 7 files listed below
4
Set Google Drive to sync this folder
Google Drive desktop app should already sync the vault folder automatically to cloud
5
Link your local code folder
In Obsidian: create note 00-PROJECT-OVERVIEW.md → paste your CLAUDE.md content here
6
Antigravity: set working directory
In Antigravity: navigate to ~/projects/study-visa-hub — this is your code root. Obsidian notes are at ~/Google Drive/AI-Builder/StudyVisaHub-Sprint/
The 7 Files to Create in Obsidian
# Create these files on Day 0. Use them every single day. 📄 00-PROJECT-OVERVIEW.md Purpose: Your CLAUDE.md copy + project goals + tech stack Update: whenever CLAUDE.md changes 📄 01-DAILY-JOURNAL.md Purpose: One entry per day — what you did, what you learned, what was hard Format: ## Day [N] - [Date] / Done: / Learned: / Stuck on: / Tomorrow: 📄 02-LEARNINGS.md Purpose: Paste end-of-session memory capture outputs here This becomes your memory.md — reusable knowledge for all future projects 📄 03-PROMPTS-THAT-WORKED.md Purpose: Every prompt that got excellent output — save it with context Format: ### [Task type] - [Date] / Prompt: / Why it worked: 📄 04-BUGS-AND-FIXES.md Purpose: Every bug you hit + how you fixed it Format: ### Bug: [description] / Error: [exact text] / Fix: [what solved it] 📄 05-DECISIONS.md Purpose: Why you made key technical or product decisions Format: ### Decision: [what] / Why: / Alternatives rejected: / Outcome: 📄 06-WINS.md Purpose: Every milestone hit. Things that surprised you positively.Read this when you feel like you're not making progress.
Daily Obsidian Routine
When
File
What to write
Time
Morning (before Claude session)
01-DAILY-JOURNAL.md
Today's target from sprint tab. Yesterday's unfinished items.
5 mins
After each major build
04-BUGS-AND-FIXES.md
Any error you hit + how you fixed it
5 mins
When Claude produces a great prompt
03-PROMPTS-THAT-WORKED.md
Paste the prompt + what it produced
2 mins
End of day (Claude memory capture)
02-LEARNINGS.md + 01-JOURNAL.md
Paste Claude's end-of-session output. Add your own thoughts.
10 mins
Deploy This Playbook to Mediamantra Website
What we're doing
You want this AI Builder Playbook HTML file hosted as additional pages on your mediamantra website — so you and your son can access it live, check things off, and take notes from any device. Here are three options depending on your mediamantra setup.
Upload the HTML file to your media library or server
WordPress Dashboard → Media → Add New → Upload AI_Builder_Playbook_v8.html
2
Create a new WordPress page
Pages → Add New → title: "AI Builder Playbook" → set template to "Full Width" or blank
3
Embed the HTML file using an iframe or HTML block
In the page editor, add an HTML block → paste: <iframe src="/wp-content/uploads/AI_Builder_Playbook_v8.html" style="width:100%;height:100vh;border:none"></iframe>
4
OR upload to Elementor/Divi as a custom template
If using a page builder: add a Code/HTML widget → paste the full HTML content directly
5
Set page to Password Protected
In Page Settings → Visibility → Password Protected → set a password for you + son only
6
Test on mobile
Open the URL on phone. The playbook should be fully functional — tabs work, checkboxes save.
Option 2 — GitHub Pages (Recommended — simplest)
# Fastest and cleanest option — no WordPress complications 1. Create new GitHub repo: mediamantra-playbook (public)
2. Upload AI_Builder_Playbook_v8.html as index.html
3. Settings → Pages → main → /root → Save
4. URL: [yourname].github.io/mediamantra-playbook # Share that URL with your son# Update by committing a new version of the HTML file# All checkboxes + notes save to browser localStorage on each device separately
Option 3 — Vercel (if you want it on a custom subdomain)
# Deploy as playbook.mediamantra.com.au 1. Create GitHub repo with the HTML file as index.html
2. Connect to Vercel → auto-deploys on every push
3. In Vercel: Settings → Domains → add: playbook.mediamantra.com.au
4. In Cloudflare (if DNS there): add CNAME record:
Name: playbook
Target: cname.vercel-dns.com
5. Wait for DNS propagation (5-30 mins)
6. Test: playbook.mediamantra.com.au loads the full playbook
Recommended: Use Option 2 Today, Option 3 Later
Fastest path
GitHub Pages takes 3 minutes. Do that now to get it live. When you're comfortable with Vercel (after SVH is deployed), move it to a proper subdomain like playbook.mediamantra.com.au or learning.mediamantra.com.au
How to Update the Playbook After Changes
# Every time we improve the playbook in Claude: 1. Download the new HTML file from Claude artifacts
2. Go to your GitHub repo (mediamantra-playbook)
3. Click index.html → Edit (pencil) → paste new content → commit
4. GitHub Pages auto-rebuilds → live within 2 minutes
5. Refresh your bookmarked URL → all changes appear
Course Gaps — All Identified and Fixed in This Version
What was missing — now added
Before this final build, the course had 12 identifiable gaps. All of them are now addressed in this tab or updated sections elsewhere in the playbook.
Hour-by-hour daily schedule, theory/practical ratio, energy management
No "what to do when stuck" protocol
Sprint → When Stuck tab
6-step unstuck protocol, escalation path, when to skip and move on
No local file → Claude Code handoff procedure
Sprint → Obsidian Setup
Exact folder paths, how Antigravity reads local files, mirroring setup
No mediamantra deployment guide
Sprint → Deploy to Mediamantra tab
3 options: WordPress, GitHub Pages, Vercel subdomain — with step-by-step
No daily standup prompt
Sprint → Day 0 Setup
Daily opening prompt for Claude — paste every morning
No rapid learning protocol
Sprint → Day 0 Setup
20-minute concept learning prompt — learn anything fast
No persistent journal in playbook
Journal & Notes tab (new)
Full note-taker with tags, search, export — saves to localStorage
No daily checklist with today's focus
Daily Checklist tab (new)
Today's tasks auto-populated from sprint, with quick-add and reset
n8n covered in theory only — no build walkthrough
Sprint Days 4 + 8
Step-by-step DHA alert build on Day 4, social automation on Day 8
Stripe mentioned but not taught
Sprint Day 9 + Technical Skills tab
Dedicated lesson slot + resource links + practical build on Day 9
No monetisation activation plan
Sprint Day 10
Launch day includes: LinkedIn post, agent email blast, social calendar
Daily Rhythm — How to Work Efficiently
The Rule
Every day is 30% theory (watching/reading) and 70% building. You learn by doing — not by watching. Every theory session must be immediately followed by applying it to SVH.
Ideal Daily Schedule
Time
Activity
Duration
Notes
8:30am
Morning standup — open Claude with daily standup prompt
10 mins
Set today's ONE priority. No more than 3 tasks total.
9:00am
Theory block — watch course / read docs
60 mins
Take notes in Obsidian as you watch. Pause and write after each section.
10:00am
Build block 1 — apply what you just learned
90 mins
One task. One chat. Paste full context. Don't switch tasks mid-session.
11:30am
Break — step away from screen
15 mins
Walk. Your brain consolidates what it learned.
11:45am
Build block 2 — second task
90 mins
New chat for new task. Paste daily standup context again.
1:15pm
Lunch break
45 mins
Don't work through lunch. You need the reset.
2:00pm
Build block 3 — third task or fix bugs from morning
90 mins
If you're debugging: paste the EXACT error text to Claude first.
3:30pm
Theory block 2 — freeCodeCamp / docs reading
45 mins
Technical reading. Not videos. Exercises if freeCodeCamp.
4:15pm
Testing + pre-deploy checks
45 mins
Test everything you built today. Run pre-deploy skill. Fix what fails.
5:00pm
Commit + end-of-session memory capture
30 mins
Commit with proper message. Run memory prompt. Update Obsidian journal.
Efficiency Rules
Rule
Why
One task per Claude session
Multi-task sessions produce worse output and waste context
Never describe an error — paste the exact text
Claude can only debug what it can read. Descriptions are always incomplete.
Commit before making any change
30 seconds of commit time saves hours of recovery time
If stuck for more than 20 minutes — use the "When Stuck" protocol
20 minutes is the limit. After that you're burning time, not learning.
Read theory in the morning when your brain is fresh
Building in the afternoon is fine with lower cognitive energy. Learning is not.
End every day with the memory capture prompt
What is not written is forgotten by morning. This is non-negotiable.
When Stuck — The 6-Step Protocol
The Limit
If you have been stuck on the same problem for more than 20 minutes, something has gone wrong with the process — not the code. Follow these steps in order.
Step
Action
Time limit
1
Open F12 → Console. Read the red text. Copy it exactly.
2 mins
2
Open a NEW Claude chat. Paste: "Error: [exact text]. File: [paste full file]. What is causing this and what is the fix?"
5 mins
3
If Claude's fix doesn't work: git checkout [last working commit] -- [filename]. Go back to working state.
5 mins
4
If you can't identify the last working commit: git log --oneline -10. Find the last "tested working" message.
3 mins
5
Start the feature again from scratch in a new Claude chat. Paste the working file + one specific task only.
—
6
If still stuck after step 5: log it in Obsidian 04-BUGS-AND-FIXES.md. Mark the task as deferred. Move to the next sprint item. Come back to it tomorrow.
—
Common Sticking Points and Their Fixes
Symptom
Likely cause
Fix
"The page is blank"
JavaScript error crashing the page before it renders
F12 → Console → read the red error → paste to Claude
"The form submits but nothing saves to Supabase"
RLS blocking the insert, or wrong table name
Supabase dashboard → Table Editor → check if row appeared. If not: check RLS policy allows insert for anon role.
"Vercel deploys but the site shows old content"
Browser cache
Ctrl+Shift+R (hard refresh). Or open in incognito.
"Claude keeps giving me the same wrong answer"
Context is stale or too long
Run /compact. Or start a new chat with the daily standup prompt.
"n8n workflow runs but nothing happens"
A node errored silently
In n8n: click each node after a run. Red nodes failed. Click the red node to see the error.
"I don't understand what Claude built"
You haven't asked it to explain
In the same chat: "Explain what you just built in plain English, section by section. I am a non-developer."
⬡Universal unstuck promptDebug
I am stuck. I have been working on this for [X] minutes. What I was trying to do: [one sentence]
What I expected to happen: [one sentence]
What actually happened: [one sentence]
Console error (exact text): [paste it or write "none"]
Last thing that worked: [describe the last working state] I am a non-developer. Diagnose this in plain English.
Tell me: (1) what is causing this, (2) the exact fix, (3) how to avoid this next time.
If the fix requires changing the code, show me the complete updated file.
Journal & Notes
Your persistent learning journal. Notes save automatically in your browser. Export anytime. Use every day — this is your Obsidian inside the playbook.
📝 Today
📚 All Notes
💡 Learnings
🐛 Bugs & Fixes
⬡ Saved Prompts
🏆 Wins
Today's Notes
Your Learning Log
Notes tagged "Learning" appear here. Use this as your memory.md equivalent — paste end-of-session captures here.
Bugs & Fixes Log
Every bug you hit + how you fixed it. This becomes your troubleshooting reference for future projects.
Saved Prompts
Prompts you tagged as worth saving. Your personal prompt library that grows with every project.
Wins 🏆
Read this when you feel like you're not making progress. Every milestone, breakthrough, and moment of "I actually got it."
Daily Checklist
Today's tasks. Tick as you go. Resets each day. Add your own tasks. Everything saves automatically.
Today's Sprint Tasks
These come from your sprint plan. Click to add to your daily checklist.