The Evolution System — Drop-in Self-Improvement Package

What's in the package

9 files. Drop them in. Run the README.

README.md 15-minute setup walkthrough. Step-by-step installation: where each file goes, which crons to activate in what order, and how to verify the system is running.

EVOLUTION-SYSTEM.md Full architecture design doc. The 4-layer system explained — Reflection Engine, Strategy Loop, Vision Council, and how they connect. Reference doc, not required reading to install.

scripts/knowledge-ingestion.sh Daily reading automation. Reads your reading queue JSON, ingests the next item, extracts key lessons, and appends structured learnings to your MEMORY.md. Requires jq for JSON handling.

scripts/system-audit.sh Weekly health check automation. Checks for crons with no recent output, stale MEMORY.md entries, and learnings that haven't been acted on. Outputs a structured audit report.

scripts/vision-council.sh Weekly strategic review automation. Fires every Sunday, reviews the week's execution output against the long-term mission, and writes structured vision output to research/vision-council/.

config/reading-queue.json Starter reading queue — 10 curated items. Paul Graham essays, Hormozi, Naval Ravikant, John Thompson. The format knowledge-ingestion.sh reads. Add your own items in the same schema.

cron-templates/evolution-crons.json All 5 cron definitions, copy-paste ready. Reflection (post-execution), knowledge ingestion (6AM daily), R&D Council (daily), Vision Council (Sunday 8PM), System Audit (Sunday). Formatted for OpenClaw cron config.

templates/reflection-block.md Structured reflection template. The exact format used in the overnight-employee cron — Rule of 3 (three learnings, three improvements, one priority action), plus a system health field.

templates/rd-council-upgrade.md Decision + Action block template. The output format for the daily R&D Council — structured so each council session produces a concrete file change or prompt update, not just analysis.

What changes

After installation, your agent operates differently.

Every significant execution ends with a structured reflection block — three learnings, three improvements, one priority action. Those learnings feed into a daily R&D Council that decides what changes, producing concrete Decision + Action blocks that update your agent's prompts and files. A daily reading cron pulls from a curated queue and extracts lessons that surface in the council. Every Sunday, a Vision Council cron reviews the full week and recalibrates the system's direction. The result: an agent whose behaviour is shaped by what it's learned, not just its initial configuration.

Honest description

Working files. Not beginner-friendly.

It is

Working bash scripts — not pseudocode, not examples
The same architecture running on a live agent right now
Real cron prompts from a production system, not placeholders
A 15-minute setup path documented in the README
Extensible — add your own reading queue items, modify prompts freely

It isn't

A tutorial — the $49 guide explains the why in depth
Plug-and-play for absolute beginners with no shell experience
A substitute for understanding what you're installing
Supported or updated on a schedule — buy what's here, use it

Real implementation

These are the actual scripts. Not samples.

knowledge-ingestion.sh reads from a live reading queue — 10 curated items ship with the package, and the format is documented so you can add your own.

system-audit.sh checks for crons with no recent output, stale MEMORY.md entries, and unacted learnings — the three failure modes I actually encountered first.

vision-council.sh fires every Sunday and outputs to research/vision-council/ — this is the actual script I run, not a cleaned-up demo version.

The cron prompts in evolution-crons.json are the real prompts from a running system. They've been revised through the same reflection loop the system teaches.

Choosing the right option

Guide vs Package vs Bundle

Guide

Deep explanation of the full 4-layer architecture. Theory + implementation. Chapter by chapter. Best if you want to understand the system well enough to extend it yourself. Includes all cron templates as appendix.

$49

AUD

Package← you are here
All 9 files ready to drop in. Scripts, configs, templates, cron definitions. 15-minute setup via README. Best if you want to implement fast and learn by doing.
$99AUD

Bundle

Guide + Package together. Everything — the deep explanation and the working files. Best if you want to understand the system fully and have it running the same day.

$129

AUD

Questions

FAQ

A working OpenClaw setup with Claude API access, and basic comfort with the command line — you'll need to place files, make scripts executable, and add cron entries to your OpenClaw config. The README assumes you've done this before. If you haven't set up OpenClaw yet, start with Guide 01.

15 minutes following README.md, assuming your OpenClaw setup is already working. This covers: placing all files in the right directories, adding the 5 cron entries to your config, verifying the scripts are executable, and running a manual test of knowledge-ingestion.sh to confirm jq is installed.

Working bash scripts. They run on my system every day. The only dependency beyond OpenClaw is jq for JSON manipulation in knowledge-ingestion.sh — install it with brew install jq on macOS. The README covers this.

Get the $49 guide instead. It explains every design decision: why 4 layers, what the Rule of 3 is and why it works, how the knowledge ingestion connects to the strategy loop, and what "self-improvement" actually means for an LLM-based agent. The package is for people who want to install first and understand later.

Yes — email [email protected] and I'll send you a bundle link at $129 AUD (saves $19 vs buying separately). You get both the full guide and all 9 package files.