Skip to content

hermesnest/midas-skill

Repository files navigation

midas.skill

"Wealth is not hiding in Bloomberg terminals or VC pitch decks. Wealth is hiding in the noise of your daily life — you've just been trained to ignore it."

License: MIT Python 3.9+ Claude Code AgentSkills

Discord


Your Slack chat leaked three business ideas this week — and you scrolled past them.
Your camera roll holds a neighborhood arbitrage thesis — and you forgot the photos existed.
Your friend group's complaints are a self-organized focus group — and nobody in the chat knows it.
Your YouTube history is 14 hours of proof you're becoming a paid specialist — and you called it "wasting time."
Your Amazon purchases are validated demand — and you think it's just shopping.

Turn Your Repeated Orders Into Gold — welcome to the noise-to-gold refinery!


Feed it any daily-life stream (Slack, photos, chats, browsing, receipts, complaints)
plus your own hunches and questions
and get a ranked list of money signals, opportunities, and immediate next actions

Supported Sources · Install · Usage · Demo · Detailed Install · 💬 Discord

中文 · Español · Deutsch · 日本語 · Русский · Português


Created by @realteamprinz | Positioning: ** Midas distills how to turn your daily noise into money.**

Supported Sources

Midas accepts any information stream you voluntarily feed it. Repetition IS the signal — if you encountered it more than once, feed it to Midas.

Source Best Lens Notes
Slack / Teams messages Demand Gap, Money Signal Team channels = vendor frustrations, budget signals, manual workarounds
WhatsApp / iMessage / WeChat chats Demand Gap, Network Path Repeated complaints from different people = unserved demand
Phone camera roll (uncurated) Arbitrage, Demand Gap Construction sites, empty storefronts, crowds, prices, signage
YouTube / TikTok watch history Skill Bridge, Behavioral Consistent topical focus = latent expertise accumulation
Browser history Skill Bridge, Money Signal Research patterns, comparison shopping, troubleshooting journeys
Purchase history / subscriptions Money Signal, Behavioral Repeat purchases = validated demand, autopilot habits
Meeting notes (internal) Money Signal, Demand Gap Budget conversations, tool complaints, process breakdowns
Email .eml / .mbox Money Signal, Network Path Commission offers, referral chains, vendor quotes
Receipts / order confirmations Money Signal, Arbitrage The most honest signal of what you actually value
Paste any text directly all 6 lenses Even a single voice-note transcript produces signals

See references/sources-guide.md for the full source hierarchy and the triangulation rules.


Install

Claude Code

Important: Claude Code looks for skills in .claude/skills/ at the git repo root. Run this in the right place.

# Install to current project (run at git repo root)
mkdir -p .claude/skills
git clone https://github.com/realteamprinz/midas-skill .claude/skills/midas-skill

# Or install globally (available in all projects)
git clone https://github.com/realteamprinz/midas-skill ~/.claude/skills/midas-skill

OpenClaw

git clone https://github.com/realteamprinz/midas-skill ~/.openclaw/workspace/skills/midas-skill

Dependencies (optional)

Midas's core skill (SKILL.md) is pure Markdown — no runtime dependencies. The Python tools under tools/ use only the standard library (Python 3.9+):

# No pip install needed — all tools use only stdlib (json, sqlite3, csv, re, argparse)
# Optional future deps are listed in requirements.txt

Python Tools

Tool What it does Usage
tools/slack_export_parser.py Parse Slack JSON/text exports into Midas input python tools/slack_export_parser.py ~/slack-export/
tools/chat_export_parser.py Parse WhatsApp/WeChat/Telegram/iMessage exports python tools/chat_export_parser.py chat.txt
tools/browser_history_parser.py Extract Chrome/Firefox/Safari history from local SQLite python tools/browser_history_parser.py --days 7
tools/evolution_manager.py Manage evolution.jsonl (list, show, stats, decay, golden) python tools/evolution_manager.py golden
tools/signal_report_validator.py Validate signal reports against Midas quality standards python tools/signal_report_validator.py report.md

Usage

In Claude Code or OpenClaw, type any of Midas's trigger phrases:

Midas, mine this                 → paste any input, get signal report
Turn this into gold              → same thing
What money signal is here        → more specific framing
点石成金                         → same, in Chinese
Midas, analyze this deal         → deal analyzer mode (see references/deal-template.md)
What am I missing                → feed a week of your own messages

Midas runs every input through the 6-lens extraction engine, cross-references against your evolution/evolution.jsonl log, and returns a ranked Signal Report with Immediate Next Actions.

The 6 Lenses

# Lens Core Question
1 Money Signal Where is money moving, stuck, or leaking?
2 Demand Gap What do people want that nobody provides?
3 Arbitrage Window Where is value being mispriced?
4 Skill Bridge What do you already know that has market value?
5 Network Path Who should be talking to whom, and what's the finder's fee?
6 Behavioral Leverage What autopilot habit is one pivot away from revenue?

Full methodology in references/signal-extraction-framework.md.


Demo

Input: "Midas, mine this — here's my last week of Slack: [73 messages pasted]"

[MIDAS SIGNAL REPORT]

Input type:       slack_messages
Volume:           73 messages across 3 channels + 4 DMs
Scan date:        2026-04-09

🟡 MONEY SIGNALS DETECTED: 7
🔴 NOISE DISCARDED:        ~65 messages (89%)

🥇 #1 — Marcus as Internal Consultant (Skill Bridge Wedge)
    Confidence: 50% | Effort: Low | Upside: $3–8k/mo side income
    Evidence:
      - Sarah DM: "if someone could solve this end-to-end for under $5k we'd probably pay it"
      - Priya DM: "I would pay for that myself"
      - 3 distinct stakeholders asked Marcus for help with 3 problems in one week
    Pattern match: Buffett — stay in circle of competence

    ⚡ Immediate next action:
       Monday 9 AM, DM Sarah with a scoped offer:
       "For the AWS tagging project — fixed fee $3,500, delivered by next Friday."

🥈 #2 — AWS Cost-Visibility One-Time Consulting (Productized)
    Confidence: 52% | Effort: Low | Upside: $5k per engagement, repeatable
    Pattern match: Thiel — small monopoly nobody takes seriously
    ...

🥉 #3 — Marcus's Shadow Spreadsheet → PM Tool (Product Play)
    Confidence: 45% | Effort: High | Upside: $5–20k/mo if productized
    ...

Full walkthrough in examples/daily-slack-mining/.

Five example runs are included:

Example Input type Result
daily-slack-mining Work Slack, 1 week 7 signals, 5 opportunities from "boring" messages
photo-roll-mining 30 phone photos, 1 week 🏆 GOLDEN at 75% — full woodworking commission practice surfaced
complaint-mining WhatsApp group, 2 weeks 🏆 GOLDEN at 78% — neighborhood roster opportunity
browsing-mining Browser + YouTube, 1 week 🏆 GOLDEN at 82% — highest confidence in example set
billionaire-pattern-match Cumulative Matches 4 user profiles against Musk/Buffett/Thiel playbooks

Detailed Install

Step 1 — Clone the repo

git clone https://github.com/realteamprinz/midas-skill
cd midas-skill

Step 2 — Place it where Claude Code looks

For Claude Code, skills live in .claude/skills/ at the git repo root (per-project) or ~/.claude/skills/ (global). Pick one:

# Per-project: first cd to the git repo you want the skill attached to
mkdir -p .claude/skills
cp -r /path/to/midas-skill .claude/skills/midas-skill

# Or global (available in every project)
mkdir -p ~/.claude/skills
cp -r /path/to/midas-skill ~/.claude/skills/midas-skill

Step 3 — Verify the skill loads

Open a Claude Code session in the repo where you installed it. Type:

What skills are available?

Midas should show up as midas-skill. If it doesn't, double-check that SKILL.md is at the root of the installed folder and the frontmatter is valid YAML.

Step 4 — Run your first mine

Paste any noisy input with a trigger phrase:

Midas, mine this — last 50 messages from my #general Slack:
[paste content]

Midas will produce your first Signal Report and append the first entry to evolution/evolution.jsonl.

Step 5 — Feed variety over time

Midas gets sharper the more varied the input. See references/sources-guide.md for the recommended weekly rotation (Slack on Mon, photos on Wed, browsing on Thu, etc.). Day 1 gives you guesses. Day 30 gives you conviction. Day 90 gives you a playbook.


Features

🔍 6-Lens Extraction Engine

Every input runs through Money Signal / Demand Gap / Arbitrage / Skill Bridge / Network Path / Behavioral Leverage. Every signal is traceable to the exact phrase, image, or data point that produced it.

🧠 Self-Learning Cumulative Intelligence

Midas does not reset between sessions. Every input appends to evolution/evolution.jsonl. New signals are cross-referenced against every prior signal. Confidence only increases through independent, cross-referenced evidence.

Evidence Confidence Midas flag
1 single-source signal 15–35% Watching
2 independent sources confirming 40–65% Working hypothesis
3+ independent sources converging 70–90% 🏆 GOLDEN — ACT
Direct market validation 85–95% 🏆 GOLDEN — ACT NOW

🎯 Famous Wealth OS Pattern Matcher

Pre-built wealth operating systems for Musk, Buffett, and Thiel. When your signals accumulate, Midas automatically checks which playbook structurally matches your situation.

"Your pattern looks like early-stage Thiel: monopoly in a niche nobody takes seriously. Thiel's playbook says: dominate the small market first."

💼 Deal Analyzer

Point Midas at any specific transaction (your own or public M&A) with "Midas, analyze this deal" and get a structured 8-part breakdown: parties, structure, source of funds, risk allocation, exit path, hidden leverage, verdict, pattern match. See references/deal-template.md.

⚡ Action, Not Analysis

Every Signal Report ends with an Immediate Next Action — a specific verb + specific noun. Never "explore options." Always "DM Dave tonight and ask how many hours/week he spends on the manual report."


Project Structure

midas-skill/
├── SKILL.md                              # Main entry
├── README.md                             # This file
├── README_ZH.md · README_ES.md · ...     # Language variants
├── LICENSE                               # MIT
├── requirements.txt                      # Python dependencies (stdlib only)
├── tools/                                # Python tooling (3.9+)
│   ├── __init__.py
│   ├── slack_export_parser.py            # Parse Slack JSON/text exports
│   ├── chat_export_parser.py             # Parse WhatsApp/WeChat/Telegram/iMessage
│   ├── browser_history_parser.py         # Extract Chrome/Firefox/Safari history
│   ├── evolution_manager.py              # Manage evolution.jsonl (list, decay, golden)
│   └── signal_report_validator.py        # Validate reports against quality standards
├── references/
│   ├── signal-extraction-framework.md    # 6-lens methodology
│   ├── noise-to-gold-pipeline.md         # 7-stage extraction pipeline
│   ├── deal-template.md                  # Deal analyzer template
│   ├── sources-guide.md                  # Input source hierarchy
│   └── famous-models/                    # Musk / Buffett / Thiel wealth OS
├── examples/                             # 5 worked examples
└── evolution/
    └── evolution.jsonl                   # Your personal cumulative intelligence log

Design Principles

  1. Your noise is your ore. Every mundane interaction contains potential wealth signals.
  2. Repetition is the primary signal. If something shows up twice in your life, it's a market.
  3. Cross-reference beats single-source. One signal is a guess. Three are a conviction.
  4. Action over analysis. Midas outputs next-steps, not essays.
  5. Compounding intelligence. Day 1 gives guesses. Day 90 gives a playbook.
  6. No moral judgment. Midas studies how money moves, not how it should move.
  7. Honest boundaries. No predictions. No guarantees. No financial advice.
  8. Public information only. Only what you voluntarily feed it.

License

MIT. See LICENSE.

Not Financial Advice

Midas does not provide investment, tax, legal, or accounting advice. Nothing in a Midas output is a recommendation to buy or sell any security. Before acting on any Midas-generated opportunity that involves material capital, consult a licensed professional in your jurisdiction.


Turn Your Repeated Orders Into Gold.

About

Midas Skill — Turn Your Repeated Orders Into Gold. A self-learning wealth extraction engine for OpenClaw / Claude Code that turns daily-life noise (Slack, photos, chats, browsing, complaints) into ranked money signals and actionable opportunities.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages