@mindrian_os/cli 1.13.0-beta.10 → 1.13.0-beta.43

package/README.md

@@ -1,622 +1,156 @@
 <div align="center">
-  <img src="https://mindrianos-jsagirs-projects.vercel.app/logo_dark.svg" alt="MindrianOS" width="160" />
+  <img src="https://mindrianos-jsagirs-projects.vercel.app/logo_dark.svg" alt="MindrianOS" width="200" />
 
   # MindrianOS
 
-  **For founders stuck on a decision they can't name.**
+  **When you are walking through a problem worth solving and you cannot yet name what is missing, MindrianOS is the thinking partner that walks beside you.**
 
-  Powered by PWS -- the practical innovation methodology developed
-  over 30+ years at Johns Hopkins University.
-  Built by [Jonathan Sagir](https://www.linkedin.com/in/jonathansagir/).
+  Powered by PWS (Problems Worth Solving), an innovation methodology built and tested through 20 years of teaching by Prof. Lawrence Aronhime.
+  Engineered by Jonathan Sagir.
 
-  [![Plugin Version](https://img.shields.io/badge/plugin-v1.10.10-blue)](https://github.com/jsagir/mindrian-os-plugin)
-  [![License](https://img.shields.io/badge/license-BSL_1.1-orange)](https://github.com/jsagir/mindrian-os-plugin/blob/main/LICENSE)
-  [![Commands](https://img.shields.io/badge/commands-71+-green)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Skills](https://img.shields.io/badge/skills-10-cyan)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Agents](https://img.shields.io/badge/agents-13-orange)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Hooks](https://img.shields.io/badge/hooks-11-red)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Edge Types](https://img.shields.io/badge/edge_types-12-yellow)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Brain Nodes](https://img.shields.io/badge/brain_nodes-32K+-purple)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Node](https://img.shields.io/badge/node-%3E%3D22.5.0-brightgreen)](https://github.com/jsagir/mindrian-os-plugin)
-  [![Surfaces](https://img.shields.io/badge/surfaces-CLI_+_Desktop_+_Cowork-brightgreen)](https://github.com/jsagir/mindrian-os-plugin)
-
-  [Website](https://mindrianos-jsagirs-projects.vercel.app) |
-  [Marketplace](https://github.com/jsagir/mindrian-marketplace) |
-  [Brain Access](https://mindrianos-jsagirs-projects.vercel.app/brain-access) |
-  [Roadmap](https://mindrianos-jsagirs-projects.vercel.app/roadmap)
+  [![Version](https://img.shields.io/badge/v1.13.0-The_Closed_Loop-1E3A6E)](CHANGELOG.md)
+  [![License](https://img.shields.io/badge/license-BSL_1.1-C8A43C)](LICENSE)
+  [![Works on](https://img.shields.io/badge/CLI_+_Desktop_+_Cowork-2D6B4A)](#three-surfaces)
 
+  [Website](https://mindrianos-jsagirs-projects.vercel.app) ·
+  [Marketplace](https://github.com/jsagir/mindrian-marketplace) ·
+  [Brain Access](https://mindrianos-jsagirs-projects.vercel.app/brain-access)
 </div>
 
 ---
 
-## What MindrianOS Actually Does (v1.10.10)
-
-You give it your documents, your proposals, your meeting transcripts, your half-formed ideas. It files them into a structured Data Room, builds a knowledge graph of relationships between them (INFORMS, CONTRADICTS, CONVERGES, INVALIDATES), and then surfaces what you cannot see on your own:
-
-- **Whitespace** -- what's missing from your thinking relative to 77 methodology frameworks (`/mos:whitespace`)
-- **Hidden connections** -- surprising cross-domain patterns between your artifacts (`/mos:find-connections`)
-- **Reverse salients** -- where your expanding system is lagging behind itself (`/mos:find-bottlenecks`)
-- **Innovation scoring** -- which domain crossings have the highest untapped potential (`/mos:score-innovation`)
-- **Convergence signals** -- themes appearing across 3+ artifacts that you haven't noticed
-- **Contradictions** -- places where your own work disagrees with itself
+## The answer first
 
-The intelligence layer queries a local SQLite graph (room.db) using targeted SQL -- not brute-force context scanning. A question like "what contradicts my market analysis?" costs ~3,500 tokens (surgical graph query + relevant artifact excerpts), not ~200,000 tokens (reading the entire room). **Up to 57x cheaper. Better answers.** (Measured per room via `/mos:scout efficiency`; see docs/CANON-PHASE-MAP.md Part 2 for the telemetry invariant.)
+> When you talk to Larry about what you are working on, MindrianOS turns the conversation into a structured room, remembers your decisions across sessions, and surfaces the contradictions you would otherwise miss.
 
-Every discovery gets filed into an **Opportunity Bank** -- a scored, ranked collection of insights you can draw from for proposals, pitches, partnerships, and pivots. The Opportunity Bank compounds: every methodology session adds to it, every meeting transcript enriches it, every new artifact creates new edges.
-
-**The result:** a researcher used MindrianOS for 2 days on a federal proposal and his team lead said "Interesting. This is helpful." A corporate VC partner asked "what does proactive mean?" three times -- then the system generated a deck where every slide answered a question she asked in the meeting. A beta tester reported a bug with 5 screenshots, and the fix shipped in under 12 hours.
+You do not learn a tool. You talk. The room takes shape underneath the conversation. Whatever you said yesterday is still working for you today.
 
 ---
 
-## Before Your First Session
+## How it works (in three pieces)
 
-### What a Room Is
+### Larry is the thinking partner
 
-A Room is not a folder-full-of-docs. It is a Living Data Room: filesystem + intelligence layer + cross-relationship scan, all local, all yours. Every time you file an artifact (a proposal, a meeting, a half-formed idea), MindrianOS scans the other sections of the Room and surfaces edges like INFORMS, CONTRADICTS, and CONVERGES, so Larry can tell you what changed in the rest of the venture. You start a Room with `/mos:new-project` and switch between Rooms with `/mos:rooms`. The Room is the venture treated as a nested system you navigate, not a repository you archive things into.
+Larry is the AI you talk to. Larry asks the questions that reframe the problem before you try to solve it, suggests the method that fits where you are, and files what you say into your room without making you stop to organize. You do not have to know the framework names. You describe what you are doing. Larry routes you.
 
-### Permissions
+### The Data Room is your venture made legible
 
-MindrianOS is read-heavy on your workspace and write-heavy only on `~/MindrianRooms/` and `./.mindrian/`, which means permission prompts fire often on a methodology session. The simplest stance is to start Claude Code with `claude --dangerously-skip-permissions` for the session, so nothing interrupts the work. If you would rather keep per-tool gating, paste the canonical 19-matcher block from `docs/settings-template.json` into `~/.claude/settings.json` and every prompt you would have seen becomes pre-approved. Both stances are valid; pick by how you like to work. Full matchers and the tradeoff write-up: see [Permissions](#permissions).
+Every conversation, every meeting, every decision lands in a folder structure organized by venture stage: the problem, the market, the solution, the team, the money, the IP, the meetings, the opportunities. You open it in your file manager. You back it up like any other folder. You own it.
 
-### Commands and Larry
+### The room surfaces what you cannot see
 
-Two ways into the same logic: type the command, or tell Larry what you want. If you know the move, `/mos:find-analogies` is the keyboard shortcut that gets you straight to the 5-stage TRIZ pipeline. If you do not yet know the framework, saying "find me a design analogy from biology for my pricing problem" to Larry in plain English gets you to the same place, with teaching along the way. Both paths point at the same underlying methodology; the command is the shortcut, the conversation is the teacher, and neither is better than the other. Larry is the pedagogical guide either way, with or without Brain access.
+Every time you add something new, the system compares it against everything already there. Larry tells you what just changed. What contradicts what. What connects to what. What is now missing. What you stopped checking weeks ago.
 
----
+You decide: APPROVE, REJECT (with a reason), or DEFER. The reason becomes part of the room. The next scan is smarter.
 
-## Quick Start
-
-### Option A: One-Line Install (recommended)
-
-```bash
-curl -sL https://raw.githubusercontent.com/jsagir/mindrian-os-plugin/main/install.sh | bash
-```
-
-Restart Claude Code. Larry starts talking.
-
-### Option B: Plugin Marketplace (if available)
-
-```bash
-claude plugin marketplace add jsagir/mindrian-marketplace
-claude plugin install mos@mindrian-marketplace
-```
-
-### Option C: Manual Clone
-
-```bash
-git clone https://github.com/jsagir/mindrian-os-plugin.git ~/.claude/plugins/mindrian-os
-cd ~/.claude/plugins/mindrian-os
-bash install.sh
-```
-
-Larry starts talking. The Room starts listening. KuzuDB builds your knowledge graph automatically. No setup required.
-
-```bash
-# Optional: install Python deps for HSI cross-artifact intelligence
-/mos:setup hsi
-```
+---
 
-### A note on permission prompts during install
+## What v1.13.0 changed (The Closed Loop)
 
-Claude Code asks you to approve each shell command it runs (clone, npm install,
-symlink creation, settings.json edit). This is expected. A full install triggers
-10 or more prompts. Approve each one. To stop being asked, pick "always allow"
-(option 2) the first time a prompt appears -- the rest of the install run will not
-re-prompt.
+Before this release, MindrianOS could file your work and surface intelligence, but the loop did not always close. You would say something, the system would react, and nothing carried forward. v1.13.0 closes the loop:
 
-If you hit a `WARN: skipping skill ...` line, that is fine: the skipped skill is
-non-critical for first-load and the install continues normally.
+- **Larry leads turn one.** The first thing you see is a conversation, not a command menu.
+- **Your first sentence becomes a room.** Type a venture sentence, get a 30-second brief and a populated room before you have to commit anything else.
+- **Tensions persist across sessions.** Larry remembers what was unresolved and brings it back when relevant.
+- **Every conversation produces an artifact.** A first session leaves you with a real room, not an empty wizard.
+- **Decisions teach the system.** Your approvals, your rejections, your reasons become part of the room's working memory.
 
-Touch-first Windows devices (Surface tablets and similar): connect an external keyboard
-before installing. The "always allow" shortcut and several install steps are keyboard-driven.
+Currently shipping as `v1.13.0-beta.19`. Final `v1.13.0` is imminent.
 
 ---
 
-## Manual Recovery
-
-If `bash install.sh` halted partway (older versions could do this; current versions warn and continue), here is how to finish the install by hand. This procedure is what a tester's Claude Code reconstructed on 2026-05-08; see `docs/testers/gary-laben/FEEDBACK.md`.
-
-**Step 1 (do this first).** Just re-run it: `bash "$HOME/.claude/plugins/mindrian-os/install.sh"`. Current versions are idempotent and no longer halt on a missing skill file.
+## What you do in a session
 
-**Step 2 (if that still does not work).** Complete it by hand. (1) Symlink the agents, (2) write the `~/.claude/settings.json` fragments (`SessionStart` hook entry, `agent: "larry-extended"`, `env.MINDRIAN_OS_ROOT` pointing at the install dir -- Step 7 of `install.sh` writes exactly these, so the simplest path is re-running the script), (3) stamp the statusline block:
+Talk. Type a command when you know the shortcut. Let Larry teach you when you do not.
 
 ```bash
-INSTALL_DIR="$HOME/.claude/plugins/mindrian-os"
-for f in "$INSTALL_DIR/agents/"*.md; do ln -sf "$f" "$HOME/.claude/agents/$(basename "$f")"; done
-bash "$INSTALL_DIR/install.sh"   # re-run: idempotently writes the settings.json fragments
-node "$INSTALL_DIR/scripts/doctor.cjs" --statusline-visibility --fix
+/mos:new-project          # tell Larry what you are exploring
+/mos:beautiful-question   # reframe the problem before solving it
+/mos:analyze-needs        # who has this problem, how badly, what they have tried
+/mos:lean-canvas          # one-page business model
+/mos:file-meeting         # paste a transcript, Larry files it
+/mos:opportunities        # what grants match this room right now
+/mos:query "what is the weakest assumption in my financial model?"
+/mos:grade                # honest assessment against real ventures
 ```
 
-PowerShell users: run the install command from CMD instead (see note below); the `ln -sf` loop has a `New-Item -ItemType SymbolicLink` equivalent if you must do it natively.
-
-**Step 3 (verify).** Inside Claude Code, run `/mos:doctor --all`. It should report all-green or name exactly what is still missing.
-
-### If the install command fails in PowerShell
-
-Open CMD instead: Start menu, type `cmd`, press Enter, then run the install command there. Several testers have had the install work in CMD when it failed in PowerShell.
-
-A note on permission prompts: 10 or more permission prompts during install is normal; see "A note on permission prompts during install" above (pick "always allow", option 2, the first time).
+You do not have to memorize these. Describe what you are trying to do. Larry routes you.
 
 ---
 
-## Permissions
+## Why the room compounds
 
-MindrianOS is read-heavy on your workspace and write-heavy only on `~/MindrianRooms/` (your rooms) and `./.mindrian/` (session state). It never writes to brain.mindrian.ai. Every `/mos:*` command respects the [Canon Part 8 Graph Boundary](docs/MINDRIAN-CANON.md#part-8---the-graph-boundary-security-constitution): your artifacts, decisions, and meetings stay local.
+Most tools get messier the more you put in. Search ranks worse. Folders bloat. The AI forgets what you told it last session. MindrianOS goes the other way.
 
-Two options for handling permission prompts:
+The mechanism is plain. Everything you say to Larry, every meeting you file, every decision you make and reason you give becomes part of your room. The room is searchable, structured, and remembered. Every NEW entry compares against everything already there.
 
-### Option 1: Nuclear -- `--dangerously-skip-permissions`
+Day one, you have a folder.
 
-Start Claude Code with `claude --dangerously-skip-permissions` and no prompts fire for the session. Appropriate for methodology workflow because the read/write surface is bounded (workspace + your rooms). Review the flag's [full warning](https://docs.claude.com/en/docs/claude-code/settings#permissions-dangerous) before using it.
+Day thirty, you have a folder that catches the contradiction between yesterday's strategy call and last week's customer interview, because nothing about either was forgotten. Larry brings back the assumption you made in week two when you are about to make a decision in week eight that depends on it. The room finds the connection between two meetings that happened a month apart that nobody remembers being related.
 
-### Option 2: Granular -- `settings.json` pre-approval
-
-Paste the canonical matcher set from [`docs/settings-template.json`](docs/settings-template.json) into `~/.claude/settings.json`. Granular per-subcommand matchers (`Bash(git diff:*)` not bare `Bash`) give fine-grained control without friction:
-
-```json
-{
-  "permissions": {
-    "allow": [
-      "Bash(git status:*)",
-      "Bash(git diff:*)",
-      "Bash(git log:*)",
-      "Bash(node bin/mindrian-tools.cjs:*)",
-      "Bash(node scripts/*.cjs:*)",
-      "Read(**)",
-      "Write(~/MindrianRooms/**)",
-      "Write(./.mindrian/**)",
-      "WebFetch(domain:api.grants.gov)",
-      "WebFetch(domain:api.tavily.com)"
-    ]
-  }
-}
-```
-
-Full list in `docs/settings-template.json` (19 matchers). Copy the whole block; Claude Code merges it with any existing `permissions.allow` entries.
-
-**When to use which**: Option 1 if you trust the surface and want zero friction. Option 2 if you run MindrianOS alongside other plugins whose permission scope you prefer to control separately.
+Nothing forgets. Everything compares. Your own past work works for you.
 
 ---
 
-## Three Ways to Use MindrianOS
-
-| Surface | Setup | Best For |
-|---------|-------|----------|
-| **Claude Code CLI** | `claude plugin install mos@mindrian-marketplace` | Full power -- hooks, scripts, Data Room, exports |
-| **Claude Desktop** | One line in `claude_desktop_config.json` | Conversational -- talk to Larry, browse your room |
-| **Cowork** | Same MCP config, shared `00_Context/` | Team ventures -- multi-user Data Room |
-
-### Claude Desktop / Cowork Setup
-
-```json
-{
-  "mcpServers": {
-    "mindrian-os": {
-      "command": "node",
-      "args": ["/path/to/MindrianOS-Plugin/bin/mindrian-mcp-server.cjs"],
-      "env": {
-        "MINDRIAN_ROOM": "/path/to/your/room"
-      }
-    }
-  }
-}
-```
-
-Restart Desktop. 6 hierarchical tools, 5 resources, 5 prompts appear. Larry works identically.
-
----
-
-## What You Get
-
-### Larry -- Your Thinking Partner
-
-Not a chatbot. A teaching agent modeled on 30+ years of classroom methodology:
-
-- **Provocative** -- "You're thinking about this as a marketplace problem. What if it's actually a logistics problem?"
-- **Structured** -- 26 PWS frameworks applied invisibly, earned through conversation
-- **Transparent** -- shows his thinking with visual traces (problem type, chosen framework, why)
-- **Honest** -- tells you what's weak in your thinking before investors do
-
-> **Larry's Thinking**
-> Problem -- Wicked (8/10 characteristics)
-> Stage -- Pre-Opportunity
-> Method -- Bono Six Hats *divergent exploration needed*
-> Chain -- Bono -> JTBD -> Market Sizing
-> *3 Brain connections - 2 cross-references*
-
-### Data Room -- Self-Organizing Intelligence
-
-| Section | What It Captures |
-|---------|-----------------|
-| problem-definition/ | The question worth answering |
-| market-analysis/ | Who has this problem and how badly |
-| solution-design/ | Your technical approach |
-| business-model/ | How you capture value |
-| competitive-analysis/ | Who else and why you're different |
-| team-execution/ | Who builds this and how |
-| legal-ip/ | Protection strategy |
-| financial-model/ | The numbers that matter |
-| opportunity-bank/ | Discovered grants and opportunities |
-| funding/ | Active funding lifecycle tracking |
-| personas/ | AI team perspective lenses |
-| meetings/ | Meeting archives with speaker intelligence |
-| team/ | Team member profiles and knowledge landscape |
-
-### Embedded Knowledge Graph (KuzuDB)
-
-Per-project graph that grows with your venture, powered by KuzuDB as the automatic backbone:
-
-```
-.md files = what's INSIDE each section (intra-section context)
-KuzuDB graph = relationships BETWEEN sections (inter-room intelligence)
-```
-
-- 12 edge types: INFORMS, CONTRADICTS, CONVERGES, ENABLES, INVALIDATES, BELONGS_TO, REASONING_INFORMS, HSI_CONNECTION, REVERSE_SALIENT, ANALOGOUS_TO, STRUCTURALLY_ISOMORPHIC, RESOLVES_VIA
-- Natural language queries: `/mos:query "What contradicts my pricing model?"`
-- Auto-updates when you file artifacts -- every filing triggers the full cascade (classify -> KuzuDB -> graph -> git)
-- Artifact IDs and pipeline provenance tracked automatically
-- Meeting and speaker nodes stored in KuzuDB for relationship discovery
-- Cross-room detection finds connections across multiple ventures
-- Zero setup, fully local -- your venture data never leaves your machine
-
-### Two Levels of Intelligence
-
-```
-Brain (remote, optional)           Room Graph (local, always on)
-Knows which tool to use next       Tracks YOUR venture's connections
-Grades against real ventures       Finds contradictions you missed
-Discovers cross-domain parallels   Remembers what you filed last week
-
-         Brain tells you HOW to think about this kind of problem
-         Room Graph tells you WHAT your own data is saying
-```
-
-### HSI + Spectral OM-HMM Pipeline
-
-Cross-artifact intelligence that finds what you missed:
-
-- **HSI (Hybrid Similarity Index)** -- Python-native computation using sklearn TF-IDF + embeddings to measure how artifacts relate across sections
-- **Spectral OM-HMM** -- Markov chain analysis of thinking-mode transitions (Seabrook & Wiskott 2022). Spectral gap scoring replaces keyword-density proxy. Artifacts with genuinely diverse thinking patterns score higher.
-- **Reverse Salient Detection** -- automatically finds where your venture's understanding lags behind its ambition (Hughes 1983), with 15% spectral bonus for integrative thinking quality
-- **3-tier similarity** -- keyword matching, sklearn embeddings, and Pinecone semantic search
-- Results stored as KuzuDB edges -- the graph gets smarter with every artifact
-- Run `/mos:setup hsi` to install Python dependencies (sklearn, numpy)
-
-### Design-by-Analogy Pipeline (NEW in v1.6.0)
-
-Cross-domain innovation discovery through structural isomorphism:
-
-```bash
-/mos:find-analogies               # Quick on-demand analogy scan
-/mos:find-analogies --brain       # Brain-enriched cross-domain search
-/mos:find-analogies --external    # External research (AskNature, patents, academic)
-```
-
-5-stage pipeline: **Decompose** (SAPPhIRE extraction) -> **Abstract** (TRIZ parameter mapping) -> **Search** (dual-mode internal + external) -> **Transfer** (correspondence tables) -> **Validate** (structural stress-test).
+## Install
 
-Built on TRIZ (39 engineering parameters, 40 inventive principles) and SAPPhIRE (7-layer functional ontology). Discovers solutions from completely different domains that share your problem's relational structure.
-
-### Parallel Agent Patterns (NEW in v1.6.0)
+### npm (one line, recommended)
 
 ```bash
-/mos:act --swarm        # 3 frameworks in parallel across highest-gap sections
-/mos:persona --parallel # 6 De Bono hats generated simultaneously
-/mos:grade --full       # 8 sections graded at once
-/mos:research --broad   # 3-angle parallel research (academic + market + competitor)
+npx @mindrian_os/cli
 ```
 
-Cross-cascade emergent discovery: parallel filings trigger HSI recomputation, finding innovation connections that serial execution would miss.
-
-### Model Routing (NEW in v1.6.0)
+### Plugin marketplace
 
 ```bash
-/mos:models             # View current profile and agent assignments
-/mos:models set budget  # Switch to budget profile (66-86% cost reduction)
-```
-
-Per-agent model selection: Opus for teaching and grading, Sonnet for structured work, Haiku for scanning. Venture-stage adaptive hints auto-select cheaper models for early exploration and Opus for investment-stage rigor.
-
-### Sentinel Intelligence (NEW in v1.6.0)
-
-```bash
-/mos:scout              # Run all sentinel tasks manually
-```
-
-Scheduled intelligence: weekly room health checks, daily grant deadline monitoring, weekly competitor watch, weekly HSI recomputation. Always-on venture monitoring without user prompting.
-
-### Git Integration (Optional)
-
-Your Room can be a GitHub repo:
-
-```bash
-/mos:rooms git-setup    # Link room to a GitHub repo
-/mos:rooms git-status   # Check sync state
-```
-
-- Every filing auto-commits and pushes (when enabled)
-- Full version history of your venture's evolution
-- Optional -- not mandatory. Works perfectly without git.
-
-### Meeting Intelligence
-
-```bash
-/mos:file-meeting                    # Paste a transcript
-/mos:file-meeting --file notes.txt   # Provide a file
-/mos:file-meeting --audio call.mp3   # Transcribe audio (Velma)
-/mos:file-meeting --latest           # Auto-fetch from Read AI
-```
-
-Larry identifies speakers (12 roles), classifies segments, files to Data Room sections with attribution, builds meeting archive (7 files), detects convergence and contradictions, tracks action items.
-
-### Opportunity Bank + Funding Room
-
-```bash
-/mos:opportunities              # Context-driven grant discovery
-/mos:funding                    # Track funding lifecycle
+claude plugin marketplace add jsagir/mindrian-marketplace
+claude plugin install mos@mindrian-marketplace
 ```
 
-Larry reads your room data (problem domain, geography, stage) and discovers relevant grants. Confirm-first UX -- review before filing. 4-stage lifecycle: Discovered > Researched > Applying > Submitted.
-
-### AI Team Personas (De Bono Six Hats)
+### Shell
 
 ```bash
-/mos:persona                    # Generate 6 perspective lenses from room data
-/mos:persona invoke black       # Get the Risk Assessor's take
+curl -sL https://raw.githubusercontent.com/jsagir/mindrian-os-plugin/main/install.sh | bash
 ```
 
-| Hat | Persona | Focus |
-|-----|---------|-------|
-| White | Data Analyst | Facts, numbers, gaps in data |
-| Red | Intuitive Advisor | Gut feelings, emotional reactions |
-| Black | Risk Assessor | What could go wrong, why this fails |
-| Yellow | Opportunity Scout | Benefits, value, upside potential |
-| Green | Creative Strategist | Alternatives, lateral thinking |
-| Blue | Process Architect | Meta-thinking, what's missing from the process |
-
-Every output carries a "perspective lens" disclaimer. These are thinking tools, not expert advice.
+Restart Claude Code. Larry starts talking.
 
-### Professional Export
+### Update or repair an install
 
 ```bash
-/mos:export thesis          # Multi-page investment thesis
-/mos:export summary         # 1-2 page executive summary
-/mos:export report          # Due diligence report with TOC
-/mos:export profile         # Professional venture profile
-/mos:export meeting-report  # Minto-structured meeting intelligence
-```
-
-De Stijl formatted PDFs generated from your Data Room content.
-
----
-
-## Connect the Brain (Optional)
-
-Without Brain, Larry teaches well. With Brain, Larry knows which frameworks work for problems like yours, grades your work against 100+ real ventures, and discovers connections between your domain and fields you'd never think to look at.
-
-**Request access:** [mindrianos-jsagirs-projects.vercel.app/brain-access](https://mindrianos-jsagirs-projects.vercel.app/brain-access)
-
-After approval, add to your `claude_desktop_config.json`:
-
-```json
-{
-  "mcpServers": {
-    "mindrian-brain": {
-      "url": "https://mindrian-brain.onrender.com/mcp",
-      "headers": {
-        "Authorization": "Bearer YOUR_API_KEY"
-      }
-    }
-  }
-}
+mindrian-os update           # marketplace + plugin update
+mindrian-os doctor --all     # diagnose drift, suggest fixes
 ```
 
-| What you're trying to do | Without Brain | With Brain |
-|--------------------------|--------------|-----------|
-| Figure out what to work on next | Larry suggests based on your room | Larry suggests based on what worked for 100+ similar ventures |
-| Know if your thinking is rigorous | General feedback | Scored and ranked against real student work |
-| Find unexpected parallels | You have to think of them | Larry surfaces connections from unrelated fields automatically |
-| Chain frameworks in the right order | Larry's best judgment | Informed by which sequences actually produced results |
-| Get honest about your weak spots | Larry points them out | Larry can show you exactly where you rank and why |
-
-**Brain-powered commands:** `/mos:suggest-next`, `/mos:find-connections`, `/mos:compare-ventures`, `/mos:deep-grade`, `/mos:research`
+A note on install prompts: Claude Code asks you to approve each shell command. 10+ prompts is normal. Pick "always allow" the first time you see one you are happy with; the rest will not re-prompt.
 
 ---
 
-## All 66 Commands
-
-### Infrastructure (always available)
-
-| Command | Purpose |
-|---------|---------|
-| `/mos:new-project` | Start a new venture project |
-| `/mos:help` | Larry recommends what to work on next |
-| `/mos:status` | View Data Room state and venture stage |
-| `/mos:room` | Manage Data Room -- view, export, launch dashboard |
-| `/mos:rooms` | Multi-room management (list, new, open, close, archive, where) + git-setup, git-status |
-| `/mos:export` | Generate professional PDFs |
-| `/mos:pipeline` | Run multi-step framework chains |
-| `/mos:setup` | Connect Brain, transcription, meeting sources, or HSI pipeline |
-| `/mos:update` | Check for and install plugin updates |
-| `/mos:radar` | Discover new platform capabilities |
-| `/mos:act` | Autonomous engine -- Brain-driven framework selection, --chain, --dry-run |
-| `/mos:snapshot` | Export 7-view SnapshotHub HTML from your Room |
-| `/mos:admin` | Hidden self-teaching admin panel |
-
-### 26 PWS Methodology Commands
-
-**Defining the Problem Space:**
-`/mos:beautiful-question` | `/mos:explore-domains` | `/mos:map-unknowns` | `/mos:diagnose` | `/mos:root-cause` | `/mos:build-knowledge`
-
-**Understanding the Market:**
-`/mos:analyze-needs` | `/mos:explore-trends` | `/mos:analyze-timing` | `/mos:macro-trends` | `/mos:user-needs` | `/mos:explore-futures`
-
-**Challenging Your Thinking:**
-`/mos:challenge-assumptions` | `/mos:validate` | `/mos:find-bottlenecks` | `/mos:dominant-designs` | `/mos:think-hats`
-
-**Building Your Solution:**
-`/mos:structure-argument` | `/mos:scenario-plan` | `/mos:analyze-systems` | `/mos:systems-thinking` | `/mos:lean-canvas` | `/mos:leadership`
-
-**Evaluating Your Venture:**
-`/mos:grade` | `/mos:build-thesis` | `/mos:score-innovation`
+## Three surfaces
 
-### Intelligence + Meeting + Funding
+MindrianOS works wherever Claude works.
 
-| Command | Purpose |
-|---------|---------|
-| `/mos:file-meeting` | File transcript (paste, --file, --audio, --latest) |
-| `/mos:opportunities` | Context-driven grant discovery |
-| `/mos:funding` | Funding lifecycle tracking |
-| `/mos:persona` | Generate AI team perspectives (De Bono) |
-| `/mos:query` | Natural language graph queries |
-
-### Brain-Powered (requires Brain API key)
-
-| Command | Purpose |
-|---------|---------|
-| `/mos:suggest-next` | Graph-informed next step |
-| `/mos:find-connections` | Cross-domain pattern discovery |
-| `/mos:compare-ventures` | Find similar ventures and extract lessons |
-| `/mos:deep-grade` | Calibrated assessment from 100+ real projects |
-| `/mos:research` | External research with semantic cross-reference |
+| Surface | What it gives you |
+|---------|-------------------|
+| **Claude Code CLI** | Full power. Hooks fire, scripts run, the room is on disk, Larry teaches with visible structure. |
+| **Claude Desktop** | Same Larry, conversational. The Data Room shows up as inline panels (dashboard, wiki, knowledge graph). |
+| **Cowork** | Same plugin, shared room. Daily briefings, persistent perspectives, multi-user. |
 
 ---
 
-## Architecture
+## The privacy line
 
-```
-MindrianOS-Plugin/
-├── .claude-plugin/plugin.json  # Plugin manifest (v1.4.1)
-├── bin/
-│   ├── mindrian-tools.cjs      # Shared CLI entry point
-│   └── mindrian-mcp-server.cjs # MCP server (Desktop/Cowork)
-├── lib/
-│   ├── core/                   # Shared modules (room-ops, state-ops, graph-ops,
-│   │                           #   meeting-ops, opportunity-ops, persona-ops,
-│   │                           #   lazygraph-ops, section-registry, nl-graph-queries)
-│   ├── mcp/                    # MCP tools, resources, prompts, Larry context
-│   └── parity/                 # CLI/MCP parity check (CI gate)
-├── mcp-server-brain/           # Brain hosting server
-├── commands/                   # 66 commands (/mos:*)
-├── skills/                     # Auto-activated intelligence (7 skills)
-├── agents/                     # 8 agents
-├── hooks/                      # 9 hooks (SessionStart, PostToolUse, Stop + 6 Powerhouse hooks)
-├── scripts/                    # 28 bash scripts + 34 CJS modules + 3 Python scripts
-│   ├── git-ops                 # Git integration (7 subcommands)
-│   ├── compute-hsi.py          # HSI dual similarity (sklearn TF-IDF + embeddings)
-│   ├── detect-reverse-salients.py  # Cross-section innovation detection
-│   ├── hsi-to-lazygraph.cjs    # SQLite edge bridge for HSI results
-│   ├── build-graph-from-sqlite.cjs  # graph.json generated from SQLite
-│   ├── cross-room-detect.cjs   # Multi-room relationship detection
-│   └── check-hsi-deps          # Python dependency checker
-├── references/                 # PWS frameworks, meeting protocols, personas
-├── templates/                  # PDF templates (De Stijl styled)
-├── dashboard/                  # Knowledge graph viewer + chat
-├── pipelines/                  # ICM stage contracts
-├── tests/                      # 8 test suites, 100+ assertions
-└── CLAUDE.md                   # Architecture (Simon + ICM + Wicked Problems)
-```
+MindrianOS reads your workspace and writes only to your rooms (default: `~/MindrianRooms/`) and to session state (`./.mindrian/`). It does not push anything to the Brain. Brain queries carry methodology questions only, never your notes, never your decisions, never your meetings.
 
-### What You Get -- By the Numbers
-
-| Component | Count |
-|-----------|-------|
-| Commands (`/mos:*`) | 66 |
-| Agents | 8 |
-| Skills (auto-loaded) | 7 |
-| Hooks | 9 |
-| MCP Tools | 64 (9 routers) |
-| KuzuDB Edge Types | 12 |
-| Pipelines | 3 |
-| Bash Scripts | 40+ |
-| CJS Modules | 20 |
-| Python Scripts (HSI pipeline) | 3 |
-
-**Three layers:**
-
-| Layer | What | Who Owns It |
-|-------|------|-------------|
-| **Plugin** | Skills, commands, agents, hooks, MCP server | This repo |
-| **Brain** | Knows which frameworks work for which problems, calibrated from real teaching | Hosted MCP (optional, API key) |
-| **Room** | Your workspace, entries, team, meetings, KuzuDB graph, exports | You -- all data stays local |
+For zero permission prompts during a session: `claude --dangerously-skip-permissions`. The read/write surface is bounded to your workspace and your rooms. For granular control, copy the matcher set from [`docs/settings-template.json`](docs/settings-template.json) into `~/.claude/settings.json`.
 
 ---
 
-## Milestones
-
-### v1.0 MVP (shipped 2026-03-22)
-One-command install. Larry talks immediately. 26 methodology commands. Data Room with 8 DD sections. De Stijl dashboard with knowledge graph. PDF export (thesis, summary, report, profile). Brain MCP integration. Self-update system.
-
-### v2.0 Meeting Intelligence (shipped 2026-03-24)
-Meeting filing pipeline (paste/file/audio + Velma transcription). Speaker identification with 12 roles. Team room with profiles and knowledge landscape. Cross-meeting intelligence (convergence, contradictions, action items). Three-layer knowledge graph with timeline mode. Minto meeting-report PDF.
-
-### v3.0 MCP Platform & Intelligence Expansion (shipped 2026-03-25)
-- **Phase 10:** Shared core library (`mindrian-tools.cjs` + `lib/core/` modules)
-- **Phase 11:** MCP server for Desktop/Cowork (6 hierarchical tools, 5 resources, 5 prompts)
-- **Phase 12:** Brain hosting with API key management
-- **Phase 13:** Opportunity Bank + Funding Room (context-driven grants, 4-stage lifecycle)
-- **Phase 14:** AI Team Personas (De Bono Six Hats from room intelligence)
-- **Phase 15:** User Knowledge Graph (embedded graph, NL queries, auto-updates)
-- **UX:** `/mos:` prefix, thinking traces, visual confirmations, room-aware status line
-
-### v4.0 Autonomous Engine & Multi-Room (shipped 2026-03-27)
-- **`/mos:act`** -- Brain-driven autonomous framework selection and execution with `--chain` and `--dry-run`
-- **`/mos:rooms`** -- Multi-room management (list, new, open, close, archive, where)
-- **`/mos:admin`** -- Hidden self-teaching admin panel
-- **Proactive intelligence persistence** -- insights survive across sessions
-
-### v5.0 Filing Pipeline, KuzuDB & HSI (shipped 2026-03-29)
-- **Phase 26 - Git Integration:** Room = GitHub repo, auto-commit/push on filing, optional (not mandatory), `git-setup` and `git-status` subcommands
-- **Phase 27 - Filing Pipeline + KuzuDB Engine:** Every filing triggers the full cascade (classify -> KuzuDB -> graph -> git). Artifact IDs, pipeline provenance, meeting/speaker KuzuDB nodes, cross-room detection, proactive intelligence persistence
-- **Phase 27.1 - HSI + Reverse Salient Pipeline:** Python-native HSI computation (sklearn TF-IDF + embeddings), Reverse Salient cross-section detection, results as KuzuDB edges, 3-tier similarity (keyword/sklearn/Pinecone)
-
-### v1.6.0 Powerhouse (shipped 2026-03-31)
-The transformation from reactive teaching partner to proactive venture intelligence engine:
-- **Model Routing:** Per-agent model selection (quality/balanced/budget/inherit) with venture-stage adaptive hints. `/mos:models` command. 60-86% cost reduction.
-- **Hook Expansion:** 6 new Claude Code hooks (PreCompact, PostCompact, FileChanged, CwdChanged, SubagentStop, TaskCompleted). Larry never loses context. External edits auto-sync. Agent results auto-file.
-- **Parallel Agents:** `--swarm`, `--parallel`, `--full`, `--broad` flags on act/persona/grade/research. 3x speed via simultaneous execution.
-- **Spectral OM-HMM:** Markov chain thinking-mode analysis replaces keyword-density scoring. Per-artifact spectral profiles. 15% breakthrough bonus for genuine integrative thinking.
-- **Design-by-Analogy:** 5-stage pipeline (Decompose/Abstract/Search/Transfer/Validate). TRIZ contradiction matrix (39x39). SAPPhIRE functional encoding. 3 new KuzuDB edge types. `/mos:find-analogies` command.
-- **Sentinel Intelligence:** `/mos:scout` for scheduled room health checks, grant deadline monitoring, competitor watch, HSI recomputation.
-- **Platform Optimization:** Prompt cache optimization, modular CLAUDE.md via @include, deep link protocol, environment variable tuning.
-- **Future-Proofing:** KAIROS-compatible room/.context/, Coordinator Mode team manifest, formal MWP specification (525 lines), moat mandate documentation.
-
-### v6.2 RoomHub + SnapshotHub (shipped 2026-03-31)
-The Room becomes a living, adaptive intelligence hub:
-- **RoomHub:** Living adaptive intelligence hub for any Room. Adaptive Room detection (venture/website/research/general) tailors the entire experience to your project type.
-- **SnapshotHub:** `/mos:snapshot` exports a 7-view standalone HTML -- Overview, Library, Narrative, Synthesis, Blueprint, Constellation, Chat -- from your Room's current state.
-- **12-Thread Constellation Graph:** Interactive knowledge graph with De Stijl colors and spectral coloring. 12 edge types rendered as a navigable constellation.
-- **Generative Fabric Chat:** Query your graph via natural language directly inside the SnapshotHub. Ask questions, get answers grounded in your room data.
-- **JTBD-Powered Contextual Discovery:** Every 3-7 turns, Larry surfaces the next command you should run based on your current state and the Jobs-To-Be-Done framework. Commands find you, not the other way around.
-
-### v1.10.0-v1.10.5 Obsidian Vault Export + Feynman-MINTO + Wiki Fix (shipped 2026-04-12 to 2026-04-14)
-- **Obsidian Vault Export:** `/mos:vault` exports your Data Room as a fully-branded Obsidian vault with wikilinks, branded footers, and a welcome doc. `/mos:vault --mode=transplant` includes the SQLite database for room-to-room bridging.
-- **Feynman-MINTO Hybrid:** Every MINTO.md reasoning file is born compressed via Feynman engine stages 1/2/4/5. Tier-1 default runs inside the Claude session (zero API key, zero per-run cost). Tier-0 fallback is deterministic.
-- **Wiki Artifact Injection Fix:** Wiki template now shows real article content when users click sections. Per-artifact 20KB cap, per-room 2MB cap. MINTO.md governing_thought as section summary.
-
-### v1.10.7 Cross-Session Scope Injection (shipped 2026-04-14)
-Stops Claude from leaking content across rooms in cross-session memory. Active Room context block injected into session-start. Sealed-room guardrails quoted in system prompt. Write-scope-check hook blocks cross-room writes at the PreToolUse level. Triggered by a witnessed cross-session leak in production.
-
-### v1.10.8 Smart Notebook Co-Pilot (shipped 2026-04-14)
-Mullins 20-section scaffold. Stakeholder node type in the graph. Graph-to-proactive-intelligence bridge. UserPromptSubmit graph-findings injection. Voice-log writer + reader. Self-update rewrite for versioned-cache model.
-
-### v1.10.9 + v1.10.10 Windows Hotfix + Mac Parity (shipped 2026-04-15)
-The cross-platform release. 10 plans, 12 commits, 2 releases in one day:
-- **BREAKING: Node.js 22.5.0 is now the minimum.** Migrated from better-sqlite3 (native bindings) to Node.js builtin node:sqlite. Eliminates the Windows native-binding failure class permanently.
-- **MOSDeckEngine skill:** YC-grade pitch deck generator using Feynman 6-stage first-principles decomposition. Ask Larry to make a deck.
-- **Brain Cypher fix (Finding I):** brain-client.cjs was sending the wrong parameter name to Brain MCP, silently breaking every Cypher-based query path. Whitespace gap detection, causal edges, graph enrichment -- all were degraded without alerting users. Now fixed.
-- **Windows self-update fix (Finding J):** Five sub-findings: python3 path resolution, directory rename on locked files, script self-overwriting, fix-never-persists loop, and a bootstrap handoff pattern. Reported by an external beta tester.
-- **Run-hook.cmd exit code (Finding F, security-adjacent):** The sealed-room write guard was silently inert on Windows for two releases. Now the guard actually fires.
-- **Cross-platform dispatch:** lib/core/platform.cjs centralizes OS detection, terminal code page handling, and hook path resolution. Session-start banner renders correctly on Windows, Mac, and Linux.
-- **Whitespace pipeline auto-install:** Python ML dependencies (numpy, scikit-learn, sentence-transformers) auto-install on first run via scripts/lib/ensure_ml_deps.py. Works on Mac stock Python.
-- **Vault export dual-mode:** --mode=vault (Obsidian-only, default) and --mode=transplant (includes .mindrian/ for room bridging).
-- **v1.10.10 same-day hotfix:** Fixed on-stop hook validation error (hookSpecificOutput is not valid on Stop hooks, now uses systemMessage). Caught within 30 minutes of v1.10.9 shipping, fixed and pushed within 12 minutes.
-- **Credits:** External beta testers for the Windows self-update report and the structured Mac environment audit that drove this release.
+## Why PWS, why Larry
 
----
-
-## Privacy
+PWS (Problems Worth Solving) is not a checklist. It is a way of thinking about ventures as wicked problems that need to be reframed before they can be solved, and that demand a working memory because nobody can hold the whole thing in their head.
 
-Everything runs locally. Your Data Room is a folder on your machine. No data leaves unless you explicitly connect Brain (optional API key) or use web research.
+Larry is the personality that delivers PWS in your terminal. The teaching is intrinsic. You do not have to know the framework names. Larry asks the question, suggests the move, shows the chain. You decide.
 
 ---
 
@@ -624,50 +158,13 @@ Everything runs locally. Your Data Room is a folder on your machine. No data lea
 
 - **Website**: [mindrianos-jsagirs-projects.vercel.app](https://mindrianos-jsagirs-projects.vercel.app)
 - **Marketplace**: [github.com/jsagir/mindrian-marketplace](https://github.com/jsagir/mindrian-marketplace)
+- **Changelog**: [CHANGELOG.md](CHANGELOG.md)
 - **Brain Access**: [Request API Key](https://mindrianos-jsagirs-projects.vercel.app/brain-access)
-- **PWS LinkedIn**: [Problems Worth Solving](https://www.linkedin.com/company/problem-solving-workspace/)
-- **PWS Methodology**: Developed over 30+ years at Johns Hopkins University
+- **PWS, Prof. Lawrence Aronhime**: [LinkedIn](https://www.linkedin.com/in/lawrence-aronhime-8363894/)
 - **Jonathan Sagir**: [LinkedIn](https://www.linkedin.com/in/jonathansagir/)
 
 ---
 
-<details>
-<summary><strong>Theoretical Foundations</strong></summary>
-
-| Source | Contribution to MindrianOS |
-|--------|---------------------------|
-| **Simon (1962)** | Architecture of Complexity -- room sections as near-decomposable subsystems |
-| **Rittel & Webber (1973)** | Wicked Problems -- the Data Room manages ventures as wicked problems |
-| **Van Clief & McDermott (2026)** | ICM -- folder structure as agentic architecture |
-| **Tetlock (2015)** | Superforecasting -- intelligence layer as Bayesian updating |
-| **Hughes (1983)** | Reverse Salients -- LazyGraph finds where venture understanding lags |
-| **Knight (1921)** | Risk vs Uncertainty -- MindrianOS navigates uncertainty |
-| **Ashby (1956)** | Law of Requisite Variety -- 26 frameworks match venture complexity |
-| **De Bono (1985)** | Six Thinking Hats -- AI personas as structured perspective lenses |
-| **Seabrook & Wiskott (2022)** | Spectral Theory of Markov Chains -- thinking-mode transition analysis for HSI |
-| **Minto (1987)** | Pyramid Principle -- SCQA + MECE per-section reasoning |
-| **Altshuller (1999)** | TRIZ -- 39 parameters, 40 inventive principles for Design-by-Analogy |
-| **Chakrabarti et al.** | SAPPhIRE -- 7-layer functional ontology for cross-domain encoding |
-
-</details>
-
----
-
 ## License
 
-**Business Source License 1.1 (BSL-1.1)**
-
-MindrianOS is source-available, not open source.
-It ships under the Business Source License 1.1 (BSL-1.1) -- a source-available license with a time-deferred conversion to an OSS license.
-
-Source is available to read, use, and contribute to. Free for personal, academic, and internal business use -- including running MindrianOS as an installed plugin for your own projects, even if those projects are commercial.
-
-Commercial use (offering MindrianOS or a derivative as a paid service to third parties) requires a commercial license. Contact: jsagir@gmail.com
-
-Converts to Apache License 2.0 on 2030-04-16.
-
-Same licensing model as MariaDB, CockroachDB, HashiCorp (Terraform), and Sentry. The rationale: the plugin distribution layer is source-available (that is the adoption channel). The proprietary intelligence layer (Brain MCP) is the moat and is never distributed.
-
-Copyright (c) 2024-2026 Jonathan Sagir. All rights reserved.
-
-For licensing inquiries: [mindrian.ai](https://mindrian.ai)
+Source-available (BSL 1.1), not open source. Copyright Jonathan Sagir and PWS / Mindrian.