@mindrian_os/cli 2.0.0-beta.3 → 2.0.0-beta.33

package/commands/whitespace.md

@@ -0,0 +1,541 @@
+---
+name: whitespace
+description: Detect whitespace gaps in the room's coverage
+help_jtbd: "Map the whitespace zones in your domain."
+body_shape: "methodology"
+hitl_shape: "F.8"
+hitl_why: "Coverage gaps are surfaced as an independent set examined in any order."
+# Phase 267.3-07, ruled in 267.3-CLASSIFICATION.md (Row 24, navigator-ruled): first delivery at commands/whitespace.md:152, the map subcommand's density grid pairing each gap with a recommended framework.
+interactive_first_reward: methodology_reframe
+serves_jtbd: ["connect-domains", "find-problem"]
+teaching: "When you suspect a gap exists in the room's coverage of a domain, /mos:whitespace runs HSI scoring across the artifact corpus to find under-explored zones. Best after the room has 20+ entries."
+# --- Phase 122 workflow-layer frontmatter ---
+kind: methodology
+frameworks: ["HSI Semantic Surprise Analysis Assistant"]
+produces: "room/opportunity-bank/whitespace/*"
+inputs: []
+autonomous_safe: true
+ui_reference: skills/ui-system/SKILL.md
+allowed-tools:
+  - Bash
+  - Read
+  - AskUserQuestion
+# --- Phase 143.3 connector frontmatter ---
+connector:
+  connects_to_spine: true
+  sensor_triggers: [SENS-06]
+  reach_id: context_block
+  sub_mode: whitespace
+  framework: "HSI Semantic Surprise Analysis Assistant"   # MUST match the existing frameworks: value
+  posture: hold
+  hierarchy_rank: 3
+  filing: fileEvidenceWithReadback
+  plan_gated: false
+  web_scope: null
+  surface: F.1
+---
+
+<!-- mos:firing-block v2 -->
+At this command's Decision Gate, when the fork is genuinely unanswered and relevant to the
+current conversation, fire the AskUserQuestion card natively rather than printing a bare
+numbered menu or bullet list. Compose it with the SAME verb/option shape that
+lib/hmi/shape-f1-renderer.cjs (renderShapeF1) produces and that lib/hmi/selector-dispatcher.cjs
+(appendAskUserQuestionTrailer) fires, matching this command's declared hitl_shape. Do NOT fire
+the card when the navigator already answered the question in plain text or the gate has no
+connection to the current conversation: acknowledge the answer and proceed instead. Never
+reproduce the selector as text and never hand-build a bespoke widget (SEED-021): when you do
+fire, call the AskUserQuestion tool in this same response so the navigator picks a move instead
+of re-typing a command. Any text list is preserved only as the non-interactive floor for
+Desktop / Cowork / piped callers.
+<!-- /mos:firing-block -->
+
+# /mos:whitespace
+
+You are Larry. This command surfaces what's MISSING in a venture's Data Room using embedding-space density estimation, TopicForest hierarchical clustering, and cross-domain literature scanning. Body shape varies by subcommand: **Shape A (Mondrian Board)** for `map`, **Shape B (Semantic Tree)** for `tree`, **Shape E (Action Report)** for `analyze`, `hypothesis`, `score`, `external`, and `discover`.
+
+**Voice rules (LOCKED):**
+- Conversational, direct, no filler. Signature openers: "Very simply...", "Here's the thing...", "One thing I've learned..."
+- NO emoji anywhere. NO "I'd be happy to help". NO "Great question!". NO sentences starting with "I".
+- Symbol vocabulary: only these 12 glyphs: &#9632; &#9660; &#9654; &#9655; |- \- &#10003; &#8226; &#9888; &#9889; &#11036; ->
+- Error pattern: 3 lines only -- What / Why: reason / Fix: /mos:command
+
+## Subcommand Routing
+
+Parse the user's input after `/mos:whitespace`. If no subcommand is given, show the **help** listing below.
+
+| Subcommand | Body Shape | Purpose |
+|------------|-----------|---------|
+| `map` | A (Mondrian Board) | Whitespace zones as density grid |
+| `analyze ZONE_ID` | E (Action Report) | Deep-dive classification + hypothesis for one zone |
+| `hypothesis ZONE_ID` | E (Action Report) | Lazy on-demand hypothesis for one zone |
+| `tree` | B (Semantic Tree) | TopicForest coverage indicators |
+| `score` | E (Action Report) | Per-artifact novelty scores ranked |
+| `external` | E (Action Report) | Cross-domain literature whitespace |
+| `discover` | E (Action Report) | Full Discovery Cycle (HSI/RS/Analogy) |
+
+## No Args: Help
+
+When `/mos:whitespace` is called with no subcommand, display:
+
+```
+-- [Room Name] -- Whitespace -- Help --
+
+  Whitespace detection -- find what's MISSING in your understanding.
+
+  Subcommands:
+  |- map                   Density map of whitespace zones (sparsest first)
+  |- analyze ZONE_ID       Deep-dive: classification + framework chain + hypothesis
+  |- hypothesis ZONE_ID    Show or generate hypothesis for a zone
+  |- tree                  TopicForest with coverage indicators per branch
+  |- score                 Per-artifact novelty scores (highest first)
+  |- external              Cross-domain literature scan via Semantic Scholar
+  \- discover              Full Discovery Cycle: HSI + RS + Analogy whitespace
+
+  > /mos:whitespace map               Start here -- see where the gaps are
+  > /mos:whitespace tree              See topic coverage at a glance
+```
+
+## Pre-flight: Room Check
+
+Before any subcommand, resolve the active room:
+
+```bash
+bash "${CLAUDE_PLUGIN_ROOT}/scripts/resolve-room"
+```
+
+If no room found, use the 3-line error format:
+
+```
+x No Data Room found
+  Why: No room under ~/MindrianRooms/ or legacy room/ in workspace
+  Fix: /mos:new-project
+```
+
+STOP.
+
+Then check that `.mindrian/` directory exists inside the room. If not:
+
+```
+x No whitespace data found
+  Why: Whitespace pipeline has not been run on this room yet
+  Fix: /mos:whitespace map
+```
+
+STOP (unless the subcommand is `map`, which will create the data).
+
+## Subcommand: map
+
+**Body Shape:** A (Mondrian Board)
+
+### Step 1: Run Pipeline
+
+Execute the whitespace-command.cjs dispatcher:
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR map
+```
+
+This ensures embeddings exist (runs compute-whitespace-embeddings.py if needed), then runs compute-whitespace-gaps.py.
+
+### Step 2: Render 4-Zone Output
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace Map -- [Venture Stage] --
+```
+
+**Zone 2 -- Content Body (Shape A: Mondrian Board):**
+
+Display whitespace zones as a grid sorted by density_score ascending (sparsest = most interesting first).
+
+```
+  Zone                    Density   Type           Nearest Frameworks
+  ws-gap-001              0.12      Ill-Defined    JTBD, Beautiful Questions
+  ws-gap-002              0.28      Well-Defined   MECE, Issue Trees
+  ws-gap-003              0.45      Un-Defined     Analogical Reasoning
+  ws-gap-004              0.67      Wicked         Systems Thinking, Causal Loop
+```
+
+Each row shows:
+- Zone ID (left-aligned)
+- Density score (0.0-1.0, lower = more sparse = bigger gap)
+- Problem type classification
+- Top 1-2 nearest frameworks (truncated to fit 80 cols)
+
+Summary line:
+```
+  Zones: [N] total  |  [X] validated  |  Sparsest: [zone_id] (density [score])
+```
+
+**Zone 3 -- Intelligence Strip** (conditional):
+Show if any zones have density < 0.2 (severe gaps):
+```
+  &#9888; [N] zones below 0.2 density -- significant knowledge gaps detected
+  &#11036; [zone_id] has no nearby artifacts -- isolated void
+```
+
+**Zone 4 -- Action Footer (NEVER omit):**
+```
+  > /mos:whitespace analyze [sparsest_zone]   Deep-dive the biggest gap
+  > /mos:whitespace tree                      See topic coverage
+  > /mos:whitespace score                     Check artifact novelty
+```
+
+## Subcommand: analyze ZONE_ID
+
+**Body Shape:** E (Action Report)
+
+### Step 1: Run Analysis
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR analyze ZONE_ID
+```
+
+If the zone ID is not found, show:
+
+```
+x Zone not found: [ZONE_ID]
+  Why: No zone with that ID in whitespace-results.json
+  Fix: /mos:whitespace map (to see available zones)
+```
+
+### Step 2: Render 4-Zone Output
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace Analyze -- [ZONE_ID] --
+```
+
+**Zone 2 -- Content Body (Shape E: Action Report):**
+
+```
+  Zone: [ZONE_ID]
+  Density: [score]
+  Problem Type: [classification]
+  Validated: [Yes/No] ([gates passed]/[total gates])
+
+  Framework Chain:
+  |- [Framework 1] -- [description or transform]
+  |- [Framework 2] -- [description or transform]
+  \- [Framework 3] -- [description or transform]
+
+  Nearest Artifacts:
+  |- [artifact_name] ([section]/)
+  \- [artifact_name] ([section]/)
+
+  Hypothesis:
+  [Full hypothesis text from interpretation-results.json, or "Not yet generated -- run /mos:whitespace hypothesis ZONE_ID"]
+```
+
+**Zone 3 -- Intelligence Strip** (conditional):
+If zone validation failed gates, show which:
+```
+  &#9888; Failed anchor gate -- gap may be at cluster periphery, not between clusters
+```
+
+**Zone 4 -- Action Footer:**
+```
+  > /mos:whitespace hypothesis [ZONE_ID]   Generate methodology-aware hypothesis
+  > /mos:whitespace map                    Back to zone overview
+  > /mos:whitespace discover               Run full Discovery Cycle
+```
+
+## Subcommand: hypothesis ZONE_ID
+
+**Body Shape:** E (Action Report)
+
+### Step 1: Check for Existing Hypothesis
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR hypothesis ZONE_ID
+```
+
+The dispatcher checks interpretation-results.json first. If a hypothesis already exists, it returns it without re-running the pipeline (lazy evaluation).
+
+### Step 2: Render 4-Zone Output
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace Hypothesis -- [ZONE_ID] --
+```
+
+**Zone 2 -- Content Body (Shape E: Action Report):**
+
+```
+  Zone: [ZONE_ID]  |  Type: [problem_type]  |  Chain: [fw1] -> [fw2] -> [fw3]
+
+  Hypothesis:
+  [Full 3-part hypothesis text]
+
+  1. What's missing and why it matters:
+     [text]
+
+  2. Framework-driven question:
+     [text]
+
+  3. Suggested next action:
+     [text]
+```
+
+**Zone 3 -- Intelligence Strip:** Omit unless zone has HIGH contradiction signal.
+
+**Zone 4 -- Action Footer:**
+```
+  > /mos:whitespace analyze [ZONE_ID]       Full zone details
+  > /mos:whitespace map                     Back to zone overview
+  > [suggested /mos: command from hypothesis part 3]
+```
+
+## Subcommand: tree
+
+**Body Shape:** B (Semantic Tree)
+
+### Step 1: Run Pipeline
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR tree
+```
+
+Ensures topic-forest.json exists (runs compute_topic_forest.py if needed), then runs label-topic-forest.cjs.
+
+### Step 2: Render 4-Zone Output
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace Tree -- Topic Coverage --
+```
+
+**Zone 2 -- Content Body (Shape B: Semantic Tree):**
+
+Display the TopicForest as an indented tree with coverage indicators per branch.
+
+```
+  &#9660; Market Dynamics                          &#10003; covered
+  |- &#9654; Customer Segments                     &#10003; 3 artifacts
+  |- &#9654; Competitive Landscape                 &#8226; 1 artifact
+  \- &#9655; Pricing Strategy                      &#11036; whitespace
+  &#9660; Technology Architecture                   &#8226; sparse
+  |- &#9654; Core Platform                         &#10003; 2 artifacts
+  \- &#9655; Integration Layer                     &#11036; whitespace
+  &#9655; Regulatory Environment                    &#11036; whitespace
+```
+
+Coverage indicators:
+- `&#10003;` = covered (2+ artifacts mapped to this branch)
+- `&#8226;` = sparse (1 artifact, needs more)
+- `&#11036;` = whitespace (0 artifacts in this topic branch)
+
+Summary line:
+```
+  Topics: [N] branches  |  &#10003; [X] covered  |  &#8226; [Y] sparse  |  &#11036; [Z] whitespace
+```
+
+**Zone 3 -- Intelligence Strip** (conditional):
+```
+  &#11036; [N] topic branches have zero coverage
+  &#9889; [topic] converges with existing room section [section]
+```
+
+**Zone 4 -- Action Footer:**
+```
+  > /mos:whitespace map                     See density scores per zone
+  > /mos:whitespace analyze [whitespace_topic]  Investigate biggest gap
+  > /mos:explore-domains                    Fill empty branches
+```
+
+## Subcommand: score
+
+**Body Shape:** E (Action Report)
+
+### Step 1: Read Scores
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR score
+```
+
+Reads whitespace-results.json and extracts artifact_novelty_scores.
+
+### Step 2: Render 4-Zone Output
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace Scores -- Novelty Ranking --
+```
+
+**Zone 2 -- Content Body (Shape E: Action Report):**
+
+```
+  Artifact                        Section               Novelty   Nearest Concept
+  customer-jobs-analysis.md       market-analysis/      0.92      JTBD Framework
+  regulatory-landscape.md         legal-ip/             0.87      Regulatory Arbitrage
+  competitive-map.md              competitive-analysis/ 0.34      Porter's Five Forces
+  problem-statement.md            problem-definition/   0.12      Beautiful Questions
+```
+
+Sorted by novelty_score descending (most novel first).
+- Novelty 0.8+ = highly novel (far from existing knowledge)
+- Novelty 0.4-0.8 = moderately novel
+- Novelty < 0.4 = well-covered territory
+
+Summary line:
+```
+  Artifacts: [N]  |  Novel (>0.8): [X]  |  Moderate: [Y]  |  Covered (<0.4): [Z]
+```
+
+**Zone 3 -- Intelligence Strip** (conditional):
+```
+  &#9889; [artifact] scores 0.9+ novelty -- genuinely unprecedented insight
+```
+
+**Zone 4 -- Action Footer:**
+```
+  > /mos:whitespace map                     See where gaps cluster
+  > /mos:whitespace analyze [top_zone]      Investigate the sparsest zone
+  > /mos:room [section_of_top_novel]        Review your most novel work
+```
+
+## Subcommand: external
+
+**Body Shape:** E (Action Report)
+
+### Step 1: Run External Pipeline
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR external
+```
+
+If the scripts are not yet installed, the dispatcher shows:
+
+```
+x External corpus not configured
+  Why: Phase 66 Plan 02 scripts not installed yet
+  Fix: /mos:whitespace external (available after Plan 02 deployment)
+```
+
+When Semantic Scholar is fully unreachable (network failure or every query rate-limited), the dispatcher shows (Phase 88.6-03):
+
+```
+x Semantic Scholar unavailable
+  Why: All queries rate-limited or network failure
+  Fix: /mos:whitespace external (retry in 60 seconds if rate-limited)
+```
+
+### Step 2: Render 4-Zone Output (when available)
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace External -- Cross-Domain --
+```
+
+**Zone 2 -- Content Body (Shape E: Action Report):**
+
+```
+  Action: External whitespace scan
+  Source: Semantic Scholar API
+
+  Cross-Domain Zones Found: [N]
+  |- [zone_id]: [description] (from: [paper_title])
+  |- [zone_id]: [description] (from: [paper_title])
+  \- [zone_id]: [description] (from: [paper_title])
+
+  Papers Analyzed: [N]
+  Top Matches:
+  |- [paper_title] ([year]) -- relevance: [score]
+  \- [paper_title] ([year]) -- relevance: [score]
+```
+
+**Zone 3 -- Intelligence Strip** (conditional, max 3 signals, Phase 88.6-03):
+```
+  warning N of M Semantic Scholar queries rate-limited (partial results shown)
+  warning N of M Semantic Scholar queries errored (partial results shown)
+  lightning External literature reveals [N] cross-domain connections
+```
+
+All three signals can appear together when partial results still surface cross-domain connections. Glyph names in backticks refer to the 12 approved glyphs from `skills/ui-system/SKILL.md` Section 3.
+
+**Zone 4 -- Action Footer:**
+```
+  > /mos:whitespace analyze [zone_id]       Investigate external zone
+  > /mos:whitespace discover                Run full Discovery Cycle
+  > /mos:research [topic]                   Deep-dive into a finding
+```
+
+### Rate-Limit Behavior
+
+Semantic Scholar free tier enforces a max of 1 request per second. When burst queries trigger 429 responses, rate-limited queries are logged and skipped; the pipeline continues with whatever papers returned. The Zone 3 warning surfaces `N of M queries rate-limited` so the user knows coverage may be partial and the cause is external throttling, not empty-result. Per-query outcomes (`ok` / `rate_limited` / `api_error` / `network_error` / `timeout` / `not_attempted`) are persisted in `.mindrian/external-papers.json` under the top-level `queries[]` array so the dispatcher shows real telemetry rather than guessing. Retrying after 60 seconds typically recovers the full set because the cache (7-day TTL) preserves successful results from the first pass.
+
+Canon Part 8 Graph Boundary: external papers are SIGNAL (public data) per canon Part 8. They flow LOCAL only (to `{roomDir}/.mindrian/external-papers.json`), never to Brain. No user-specific strings are ever sent to Semantic Scholar or Brain in this pipeline. The query keywords are extracted from local artifact titles and framework names, sent to the public Semantic Scholar API over HTTPS, and the returned abstracts stay on disk in the room.
+
+## Subcommand: discover
+
+**Body Shape:** E (Action Report)
+
+### Step 1: Run Discovery Cycle
+
+```bash
+node "${CLAUDE_PLUGIN_ROOT}/scripts/whitespace-command.cjs" ROOM_DIR discover
+```
+
+Runs discovery-cycle.cjs with `--steps all`. This chains HSI, RS, and Analogy whitespace detection.
+
+### Step 2: Render 4-Zone Output
+
+**Zone 1 -- Header Panel:**
+```
+-- [Room Name] -- Whitespace Discover -- Full Cycle --
+```
+
+**Zone 2 -- Content Body (Shape E: Action Report):**
+
+```
+  Action: Discovery Cycle
+  Steps: HSI -> RS -> Analogy
+
+  Results:
+  |- HSI Whitespace:     [N] zones (between connected artifacts)
+  |- RS Whitespace:      [N] zones (downstream of bottlenecks)
+  \- Analogy Whitespace: [N] zones (unmapped transfer zones)
+
+  Total Zones: [N]
+  Validated: [X] / [N]
+
+  Top Discoveries:
+  |- [zone_id] [gap_signal] -- [hypothesis preview, 80 chars]
+  |- [zone_id] [gap_signal] -- [hypothesis preview, 80 chars]
+  \- [zone_id] [gap_signal] -- [hypothesis preview, 80 chars]
+```
+
+**Zone 3 -- Intelligence Strip** (conditional):
+```
+  &#9889; Discovery found [N] strong-signal zones across [X] pipeline sources
+  &#9888; [N] zones failed validation -- may be noise
+```
+
+**Zone 4 -- Action Footer:**
+```
+  > /mos:whitespace map                     Visualize all zones
+  > /mos:whitespace analyze [top_zone]      Deep-dive the strongest signal
+  > /mos:whitespace score                   Check artifact novelty
+```
+
+## Error Handling
+
+All errors use the 3-line pattern (per D-24):
+
+```
+x [What failed]
+  Why: [Specific reason]
+  Fix: [One resolving command]
+```
+
+Common errors:
+
+- **No Python installed:** `x Python 3 not found / Why: Whitespace pipeline requires Python 3.8+ / Fix: Install Python 3 and retry`
+- **No embeddings:** `x No whitespace embeddings / Why: Pipeline needs to embed room artifacts first / Fix: /mos:whitespace map (runs embedder automatically)`
+- **Script timeout:** `x Pipeline timed out / Why: Room has too many artifacts for default timeout / Fix: Run with fewer sections or contact support`
+
+## Cross-Surface Adaptation
+
+- **CLI:** Full power. Scripts run via Bash, output formatted for terminal.
+- **Desktop:** Larry describes the whitespace findings conversationally. Numbers and zone IDs still shown but in natural language context.
+- **Cowork:** Same as CLI. Zone data can be shared via 00_Context/ for team visibility.

package/commands/wiki.md

@@ -0,0 +1,96 @@
+---
+name: wiki
+description: Open the Data Room wiki of room sections
+help_jtbd: "Open your room as a localhost wiki, hyperlinks via graph."
+body_shape: D (Document View)
+hitl_shape: "F.1"
+hitl_why: "The wiki view offers one next move on which article to open."
+# Phase 267.3-07, ruled in 267.3-CLASSIFICATION.md (Row 25): first delivery at commands/wiki.md:48, the bare-invocation live Express server rendering room content, ephemeral, nothing persisted unless --export.
+interactive_first_reward: "--none (diagnostic surface)"
+serves_jtbd: ["audit-room", "prepare-pitch"]
+teaching: "When you want to read the Data Room as linked wiki pages, /mos:wiki opens the wiki view. Section by section, with cross-references rendered as hyperlinks."
+ui_reference: skills/ui-system/SKILL.md
+allowed-tools:
+  - Bash
+  - Read
+  - AskUserQuestion
+# --- Phase 172-16 CIRS R1 WIRE (Canon Part 11; navigator-directed 2026-06-23) ---
+connector:
+  connects_to_spine: true
+  sensor_triggers: []
+  reach_id: context_block
+  sub_mode: room-wiki
+  framework: null
+  posture: hold
+  hierarchy_rank: 12
+  filing: memory_event_only
+  plan_gated: false
+  web_scope: null
+  surface: F.1
+---
+
+<!-- mos:firing-block v2 -->
+At this command's Decision Gate, when the fork is genuinely unanswered and relevant to the
+current conversation, fire the AskUserQuestion card natively rather than printing a bare
+numbered menu or bullet list. Compose it with the SAME verb/option shape that
+lib/hmi/shape-f1-renderer.cjs (renderShapeF1) produces and that lib/hmi/selector-dispatcher.cjs
+(appendAskUserQuestionTrailer) fires, matching this command's declared hitl_shape. Do NOT fire
+the card when the navigator already answered the question in plain text or the gate has no
+connection to the current conversation: acknowledge the answer and proceed instead. Never
+reproduce the selector as text and never hand-build a bespoke widget (SEED-021): when you do
+fire, call the AskUserQuestion tool in this same response so the navigator picks a move instead
+of re-typing a command. Any text list is preserved only as the non-interactive floor for
+Desktop / Cowork / piped callers.
+<!-- /mos:firing-block -->
+# /mos:wiki
+
+Open the localhost wiki dashboard for the current Data Room.
+
+## What it does
+Launches a local Express server that renders room/ .md files as Wikipedia-style pages with:
+- Room Home dashboard (landing page) + knowledge graph tab
+- BlockNote editing surface with direct save-to-markdown
+- Per-article PDF and Word (DOCX) export
+- Section navigation sidebar (collapsible folder tree)
+- Table of contents per page
+- Infobox from YAML frontmatter
+- [[wikilink]] cross-references as clickable hyperlinks
+- SQLite graph edges as "See also" and backlinks
+- Full-text search across all pages
+- Dark/light mode toggle
+- De Stijl design
+
+## Usage
+
+```bash
+# Open the wiki locally
+bash "${CLAUDE_PLUGIN_ROOT}/scripts/serve-wiki"
+
+# Export as static HTML for sharing
+bash "${CLAUDE_PLUGIN_ROOT}/scripts/serve-wiki" --export
+```
+
+## Flags
+
+- **No flags** - Opens wiki in browser at localhost:8421
+- **`--export`** - Generates a static HTML bundle in `export/wiki/` that can be:
+  - Deployed to Render or Vercel for team sharing
+  - Sent as a zip file
+  - Hosted anywhere as static files
+
+## Sharing Your Data Room
+
+Want to share with teammates or investors?
+
+1. Run `/mos:wiki --export` to generate static HTML
+2. Deploy to Vercel: `cd export/wiki && npx vercel`
+3. Or deploy to Render: push `export/wiki/` to a repo with a `render.yaml`
+4. Or just zip and send: `zip -r data-room.zip export/wiki/`
+
+**Privacy note:** The exported HTML contains your room content. Share only what you intend to share. MindrianOS does not host or access your data.
+
+## Notes
+- Port: 8421 (auto-increments if busy, range 8421-8430)
+- Direct save: editing an article and clicking Save rewrites its .md file in place. There is no confirmation dialog and no conflict check, so a concurrent external edit (your IDE, another Claude Code session) can be overwritten.
+- Graph dashboard remains at port 8420 via /mos:visualize
+- Chat panel is present (stub - full Larry integration coming)

package/data/ROOM.md

@@ -0,0 +1,47 @@
+---
+type: directory-identity
+name: data
+section: data
+purpose: Plugin-local data artifacts -- the generated command registry + Brain framework-name snapshot it validates against, the hand-maintained Brain Context Packet schema, and the hand-maintained deployment-surface manifest.
+founding_phase: 122
+phase: 123
+milestone: v1.13.0-beta.13
+canon_parts: [5, 6, 7, 8]
+created: 2026-05-12
+updated: 2026-05-13
+---
+
+# data/
+
+`data/` holds plugin-local data artifacts in two layers:
+
+1. **Generated artifacts** (Phase 122 -- the Workflow Layer): `command-registry.json` and `framework-names.json` are regenerated from source by `scripts/build-command-registry.cjs`. Nothing here is hand-edited.
+2. **Hand-maintained artifacts** (Phase 110, Phase 123): `brain-packet-schema.json` is the Brain Context Packet wire schema; `deployment-surfaces.json` is the deployment-surface manifest. Both are validated, not generated.
+
+## Files in this section
+
+| File | Generated by | Purpose |
+|------|--------------|---------|
+| `command-registry.json` | `scripts/build-command-registry.cjs` | The framework <-> command registry: `{ ontology_ref, generated_note, commands[], framework_index, curated_chains[] }`. Built from the `frameworks:` / `kind:` / `produces:` / `inputs:` / `autonomous_safe:` keys in every `commands/*.md` frontmatter. The resolver (`lib/workflow/command-resolver.cjs`) reads only this file at runtime. DO NOT hand-edit -- the pre-commit hook and the Feynman runner reject a stale registry. Regenerate: `node scripts/build-command-registry.cjs`. |
+| `framework-names.json` | `scripts/build-command-registry.cjs --refresh-names` | The committed snapshot of the FEEDS_INTO-linked Brain `:Framework` names (the ~105-name traversable slice, NOT all 748 -- junk like "Amazon" / "Charles Kirschbaum" is excluded) PLUS a small `curated_extras` whitelist of legitimate `:Framework` nodes not yet FEEDS_INTO-linked. The registry validates every `frameworks:` entry against `framework_names UNION curated_extras`; an unresolvable name fails the build. Refresh: `node scripts/build-command-registry.cjs --refresh-names` (a read-only build-time `brain.query`, never at runtime). |
+| `brain-packet-schema.json` | `scripts/build-brain-packet-schema.cjs` (validated, not generated -- hand-maintained source of truth) | The Brain Context Packet wire format (draft 2020-12 JSON Schema): `$defs` per all 12 closed-vocabulary Brain jobs (each carrying an `in` and an `out` shape), shared `$defs` for `LocalGraphSummary` / `BankedOpportunities` / the `privacy_mode` enum / the `packet_version` const / the `origin` enum / the 4 safe-projection shapes, `additionalProperties: false` on every object node (the Canon Part 8 leak-prevention teeth). `lib/core/brain-client.cjs::sendPacket` (Phase 110-03) compiles it once and validates every outbound packet and every Brain response against the job's `in`/`out`. DO NOT hand-edit carelessly -- the pre-commit hook (Phase 110-04) and the Feynman runner reject a malformed schema / a job missing `in`+`out` / an unknown-job `$def` / a missing `additionalProperties: false`. Re-validate: `node scripts/build-brain-packet-schema.cjs --check`. Bound to Phase 110; canon parts 8 + 9. |
+| `deployment-surfaces.json` | hand-maintained (NOT generated, NO `--check` tripwire) | The static deployment-surface manifest read by `scripts/session-start` (for the on-version-change owned-surface reconcile) and `scripts/doctor.cjs` class J (Plan 123-03; for drift flagging). Schema (`{ schema_version, surfaces:[{id, path, owner, topology_scope, check_kind, expected, reconcile, remediation, path_within_file?}, ...] }`) is the D-07 contract from `.planning/phases/123-install-lifecycle-harness/123-CONTEXT.md`, extended in Plan 123-03 with an optional `path_within_file` (dot-path) field that tells the `check_kind: exact-value` reader to JSON-parse `path` and compare the value at `path_within_file` against `expected` rather than substring-matching the whole file. Six entries cover: the statusline dispatch shim, the `settings.json` statusLine command (uses `path_within_file: "statusLine.command"` per the live-dev-box bug -- whole-file substring match was producing `ok:false` against a healthy settings.json), `~/.mindrian-last-version`, the install-state record (self, observed-only), the plugin-bin `$PATH` entry (observed-only), and the dev-clone pre-commit hook (`topology_scope: dev-clone` -- skipped on a user box). Paths use literal `$HOME` / `<active_root>` / `<dev_clone_root>` tokens (no absolute paths). New surface = one JSON entry, no code change. Bound to Phase 123 plans 02+03 (HARNESS-123-06 + HARNESS-123-08); canon parts 5 + 6. |
+
+## Canon Part 8 boundary (plugin-local, validated AGAINST the Brain, never written back)
+
+`command-registry.json` is plugin-local product surface -- commands NEVER enter the Brain. `framework-names.json` is the one Brain-derived artifact in this loop, and it carries only generic `:Framework` node names (a methodology handle, never user content). The `--refresh-names` query (`MATCH (f:Framework) WHERE (f)-[:FEEDS_INTO]-() RETURN f.name`) is read-only; there is no write Cypher anywhere in the generator, and nothing here is queried at runtime.
+
+## Decision #15 compliance
+
+Per `CLAUDE.md` Decision #15, every directory carries a ROOM.md identity file. `data/` was founded by Phase 122 and extended by Phase 110 (`brain-packet-schema.json`) and Phase 123 (`deployment-surfaces.json`); subsequent additions update the "Files in this section" table. No MINTO.md required at this level -- the `.room-root` cascade scope is `room/`, not `data/`.
+
+## Cross-references
+
+- Generator: `scripts/build-command-registry.cjs`
+- Schema validator (Phase 110-01): `scripts/build-brain-packet-schema.cjs`
+- Resolver (runtime consumer): `lib/workflow/command-resolver.cjs` (Phase 122-03)
+- Manifest walker (runtime consumer): `scripts/session-start` (Phase 123-02 -- on-version-change reconcile); `scripts/doctor.cjs` class J (Phase 123-03 -- drift flagging)
+- Frontmatter contract: `docs/COMMAND-FRONTMATTER.md`
+- Spec: `.planning/WORKFLOW-LAYER-SPEC.md`
+- Canon: `docs/MINDRIAN-CANON.md` Parts 5, 6, 7, 8 (and Part 9 via `brain-packet-schema.json`)
+- Sibling identity reference: `lib/workflow/ROOM.md` (Phase 122)