  ____  _     _     _     _   _   _
 |  _ \(_)___| |__ (_)___| |_| |_(_)_ __   __ _
 | |_) | / __| '_ \| / __| __| __| | '_ \ / _` |
 |  __/| \__ \ | | | \__ \ |_| |_| | | | | (_| |
 |_|   |_|___/_| |_|_|___/\__|\__|_|_| |_|\__, |
                                           |___/

AI-Accessible Company Brain · Bootstrap Protocol · Live Dashboard

🧠 Structured Knowledge

Every file has timestamp headers, skills layers, and automated updates — your AI reads before every action.

⚡ 9-Step Bootstrap

Drop FRAMEWORK.html, say the trigger phrase, and Claude builds your entire brain from scratch.

📊 Live Dashboard

MRR, ARR, churn, tasks, proactive suggestions — auto-refreshes every 5 minutes with dark mode.

🔄 Task Queue Runner

Cron-driven Node.js script processes pending tasks, falls back to inbox queuing if Claude CLI is unavailable.

📋 Setup Bootstrap Protocol

Check Context

Agent reads FRAMEWORK.html, checks for existing _brain/index.md, and assesses project state.

Knowledge Drop

Creates _brain/inbox/, prompts user to drop documents, reads and categorizes them.

Smart Interview

One question at a time — agent identifies gaps in business, sales, marketing, dev, ops.

Gap Questions

Agent asks targeted questions to fill missing knowledge areas before building structure.

Structure Plan

Proposes folder layout and file map based on extracted knowledge and interview answers.

Build Brain

Creates all brain files with real content, timestamp headers, and cross-references.

Update CLAUDE.md

Adds brain path, trigger phrase, and protocol reference to project instructions.

Propose Dashboard

Generates dashboard.html with live metrics, scheduled tasks, and dark mode toggle.

Validate

Agent answers test questions from brain files to verify completeness and accuracy.

🧭 Philosophy

Principle	Description
AI-First	Every file is structured for machine reading — headers, timestamps, clear sections.
Self-Healing	Skills auto-update when user corrects agent behavior; cron jobs keep metrics fresh.
Idle-Only Ops	Proactive scanning only runs when no human commits detected in the last hour.
Graceful Fallback	No Claude CLI? Tasks queue to inbox for next session. No dashboard? Skip silently.

📂 Folder Structure

_brain/
- index.md
- core/
  - business.md
  - product.md
  - team.md
- sales/
- marketing/
- development/
- operations/
- skills/
  - communication.md
  - coding.md
  - process.md
- inbox/
- tasks/
  - queue.json
- dashboard.html

🗺️ File Map

Path	Purpose
_brain/index.md	Root index with links to all sections, last updated timestamp
_brain/core/business.md	Business model, revenue streams, MRR/ARR targets
_brain/skills/communication.md	Tone rules, response patterns, auto-update on corrections
_brain/tasks/queue.json	Pending tasks with schedule, batch size, timeout config
_brain/dashboard.html	Live metrics dashboard with auto-refresh and dark mode

🔀 Information Routing

Source	Destination	Trigger
Inbox documents	_brain/core/, _brain/sales/	Knowledge drop processing
User correction ("always do X")	_brain/skills/[area].md	Real-time update with timestamp
Queue runner task	_brain/tasks/queue.json → execution	Cron every 15 min
Dashboard metrics	_brain/dashboard.html	Auto-refresh every 5 min

🤖 AI Protocol

TRIGGER_PHRASE: "initialize persistia brain"
STEPS:
  1. Check for _brain/index.md
  2. Create _brain/inbox/ if missing
  3. Prompt user to drop knowledge documents
  4. Read inbox files, extract key info
  5. Conduct smart interview (one question at a time)
  6. Identify gaps and propose structure
  7. Build all brain files with real content
  8. Update CLAUDE.md with brain path
  9. Propose dashboard.html
 10. Validate with test questions

📝 Timestamp Header Pattern

---
LAST_UPDATED: 2026-06-08T14:30:00Z
NEXT_UPDATE: 2026-06-09T14:30:00Z
---
# Business Overview
...

🎯 Skills Layer

Area	File	Auto-Update Trigger
Communication	_brain/skills/communication.md	"always say X", "never do Y"
Coding	_brain/skills/coding.md	"from now on, use Z pattern"
Process	_brain/skills/process.md	"always check X before Y"

⏰ Auto-Update Schedule

Task	Schedule	Fallback
Queue runner	Every 15 min	Inbox queuing if no Claude CLI
Dashboard refresh	Every 5 min	Client-side countdown timer
Proactive check	Idle only (no human commits in 1h)	Defer to next cycle
Skills update	On user correction	Immediate timestamp + change note

💡 Update Prompts

📥 Knowledge Drop: "Drop your business documents into _brain/inbox/ — I'll read, categorize, and archive them."

🔄 Skills Update: "Say 'always do X' or 'never do Y' and I'll update the relevant skills file immediately."

📊 Dashboard: "Open _brain/dashboard.html to see live metrics, scheduled tasks, and proactive suggestions."

📋 Replication Steps

Drop FRAMEWORK.html in your project root.
Say "initialize persistia brain" to Claude.
Agent checks context, creates inbox, prompts for documents.
Drop knowledge files → agent reads, categorizes, archives.
Answer smart interview questions (one at a time).
Agent builds full brain structure with real content.
Agent updates CLAUDE.md and sets up cron jobs.
Dashboard goes live — metrics auto-refresh every 5 min.
Brain is operational. Skills auto-update on corrections.