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

package/skills/operator/SKILL.md

@@ -0,0 +1,167 @@
+---
+name: operator
+description: "Show or manually set the conversation operator (JUST_TALK / EXPLORE_CAPTURE / BUILD_ROOM / METHODOLOGY / DECISION_GATE): the per-room state machine that governs how Larry renders responses"
+license: BSL-1.1. See LICENSE for complete terms (Business Source License 1.1, Change Date 2030-04-16 to Apache License 2.0).
+compatibility: Requires Claude Code (or a host implementing disable-model-invocation semantics); Tier-1 hook mechanics referenced in this skill.
+help_jtbd: "See or change the conversation operator state."
+argument-hint: "[history] [set <op>] [reset] [--json]"
+body_shape: E (Action Report)
+hitl_shape: "F.1"
+hitl_why: "The operator view offers one next move on the current state."
+# Phase 267.3-07, ruled in 267.3-CLASSIFICATION.md (Row 3): first delivery at commands/operator.md:78, the bare-invocation state readout of the conversation-operator's own current value and history.
+interactive_first_reward: "--none (diagnostic surface)"
+body_shape_detail: current state + last 5 history (default), full history (history subcommand), Shape F.1 picker (set subcommand), Shape F.4 confirmation (reset subcommand)
+serves_jtbd: ["explore"]
+teaching: "When you want to see or set how Larry should render right now, /mos:operator shows the conversation state machine: JUST_TALK, EXPLORE_CAPTURE, BUILD_ROOM, METHODOLOGY, or DECISION_GATE."
+allowed-tools: Bash Read AskUserQuestion
+disable-model-invocation: false
+# --- Phase 143.3 connector frontmatter ---
+connector:
+  connects_to_spine: true
+  sensor_triggers: []
+  reach_id: context_block
+  sub_mode: operator
+  framework: null
+  posture: hold
+  hierarchy_rank: 36
+  filing: none
+  plan_gated: false
+  web_scope: null
+---
+
+# /mos:operator
+
+Inspect or manually set the conversation operator for the active room. The operator is the state primitive that governs Larry's rendering: JUST_TALK suppresses 4-zone output, BUILD_ROOM emits full Shape E, METHODOLOGY suppresses spontaneous Zone 4 footers, DECISION_GATE locks Shape F.x.
+
+The operator state file lives at `<roomDir>/.mindrian/conversation-operator.json` (per-room, never global). Hooks (Phase 99-04) keep it fresh automatically as you type, invoke tools, and end sessions. Most users never need this command directly -- the heuristic classifier (Phase 99-02) handles transitions silently. Use `/mos:operator` when:
+
+- The poller (Phase 95.1 class F) reports the operator looks wrong
+- You want to inspect what Larry currently believes the operator is
+- You want to manually override (e.g., force JUST_TALK to stop filing for the rest of the session)
+- You are debugging the state machine
+
+## Step 1: Parse the user's intent
+
+Look at the invocation:
+
+- `/mos:operator` (no args) -> show current state + last 5 history entries (Shape E)
+- `/mos:operator history` -> show full history up to 50 entries (Shape E)
+- `/mos:operator set <op>` -> render Shape F.1 picker; if the user selected an operator inline (e.g., `set BUILD_ROOM`), perform the transition immediately and re-render Shape E showing the new state
+- `/mos:operator reset` -> render Shape F.4 confirmation; on user confirmation, transition to JUST_TALK
+- `/mos:operator --json` -> machine-readable output (for hooks / regression tests)
+
+Combine `--json` with any subcommand: `/mos:operator history --json`.
+
+## Step 2: Execute
+
+Run via Bash:
+
+```
+node "${MINDRIAN_OS_ROOT:-${CLAUDE_PLUGIN_ROOT:?MindrianOS install root not found. Set MINDRIAN_OS_ROOT (see lib/core/active-plugin-root.cjs) or run from Claude Code.}}/scripts/operator-command.cjs" $@
+```
+
+The script does the work:
+
+1. Resolves the active room from `~/MindrianRooms/.rooms/registry.json`
+2. Loads `<roomDir>/.mindrian/conversation-operator.json` via `lib/conversation/operator.cjs.getCurrent`
+3. Branches on subcommand: render Shape E, render Shape F.1, render Shape F.4, or perform a transition
+4. For `set` subcommand: validates the requested operator against the 7 transition rules; rejects invalid transitions with a 3-line stderr per Canon Part 3 Rule 2 and exits non-zero
+
+## Step 3: Render the output
+
+The script outputs a 4-zone Shape E (Action Report) per `skills/ui-system/SKILL.md`. Display the script's stdout directly. Do not re-format. Do not strip ANSI color codes.
+
+## Example output (default -- show current state)
+
+```
+-- mindrianos -- operator -- BUILD_ROOM --
+
+  ■ Current               BUILD_ROOM
+     entered  2026-05-01T10:42:00Z (3m ago)
+     previous EXPLORE_CAPTURE
+     context  active_section=research, methodology=null
+
+  ■ Last 5 history
+     ├─ JUST_TALK         2026-05-01T10:30:00Z  trigger=session_start
+     ├─ EXPLORE_CAPTURE   2026-05-01T10:35:00Z  trigger=user_message
+     └─ BUILD_ROOM        2026-05-01T10:42:00Z  trigger=mos_command
+
+  Summary: total transitions=3, history_used=3/50
+
+  ▶ /mos:operator history     # full history
+  ▷ /mos:operator set <op>    # manual transition
+  ▷ /mos:operator reset       # return to JUST_TALK
+```
+
+## Example output (history subcommand)
+
+```
+-- mindrianos -- operator -- BUILD_ROOM --
+
+  ■ Full history (3/50 entries)
+     ├─ JUST_TALK         2026-05-01T10:30:00Z  trigger=session_start          from=null
+     ├─ EXPLORE_CAPTURE   2026-05-01T10:35:00Z  trigger=user_message           from=JUST_TALK
+     └─ BUILD_ROOM        2026-05-01T10:42:00Z  trigger=mos_command            from=EXPLORE_CAPTURE
+
+  Summary: total transitions=3, history_used=3/50, oldest=2026-05-01T10:30:00Z
+
+  ▶ /mos:operator             # current state
+  ▷ /mos:operator set <op>    # manual transition
+  ▷ /mos:operator reset       # return to JUST_TALK
+```
+
+## Example output (set subcommand -- Shape F.1 picker)
+
+```
+-- mindrianos -- operator -- set --
+
+  ■ Manual operator transition
+     current: BUILD_ROOM
+
+  [F.1 Next Move]
+   ▶ JUST_TALK
+   ▷ EXPLORE_CAPTURE
+   ▷ BUILD_ROOM         (current -- selecting this is a no-op)
+   ▷ METHODOLOGY
+   ▷ DECISION_GATE
+   ▷ Free-Text
+
+  ▶ /mos:operator             # cancel and re-show state
+  ▷ /mos:operator history     # see full history first
+```
+
+## Example output (reset subcommand -- Shape F.4 confirm)
+
+```
+-- mindrianos -- operator -- reset --
+
+  ■ Reset operator to JUST_TALK
+     current: BUILD_ROOM
+     this discards your active filing context (active_section=research)
+
+  [F.4 Confirm Reset]
+   ▶ Confirm reset to JUST_TALK
+   ▷ Cancel
+
+  ▶ /mos:operator             # show current state
+  ▷ /mos:operator history     # see full history
+```
+
+## Note on Shape F.1 + F.4 deferral
+
+Per Phase 95.1-04 D-19 deferral pattern, the F.1 picker and F.4 confirmation render as STRUCTURAL marker blocks in stdout. Larry handles conversational selection: when the user types "set to METHODOLOGY" or "yes, reset" in natural language, Larry interprets and re-invokes the command with the explicit verb (e.g., `/mos:operator set METHODOLOGY`).
+
+Phase 88.2 (`uiux-selector-block`) will replace the marker block with the canonical AskUserQuestion primitive. See `.planning/phases/99-conversation-operator-state-machine/operator-shape-f1-deferred.md` for the deferral note + re-trigger condition.
+
+## Voice rules
+
+When Larry surfaces the output conversationally:
+
+- "You're currently in BUILD_ROOM. Filing is live; every response ends with the 4-zone footer."
+- "Want to override? `/mos:operator set <op>` lets you pick from the five canonical operators."
+- "If the operator looks wrong, `/mos:doctor --ui-compliance` will tell you whether the renderer is honoring it correctly."
+
+NEVER:
+- Apologize for the operator state being what it is. The state is the truth.
+- Suggest the user "should" be in a particular operator. The user picks.
+- Re-show the example output blocks above when speaking conversationally; just describe what they would see.

package/skills/opportunities/SKILL.md

@@ -0,0 +1,171 @@
+---
+name: opportunities
+description: Discover and manage grants in the Opportunity Bank
+license: BSL-1.1. See LICENSE for complete terms (Business Source License 1.1, Change Date 2030-04-16 to Apache License 2.0).
+help_jtbd: "Browse banked opportunities by HSI score."
+argument-hint: "[list|add|scan]"
+body_shape: E (Action Report)
+hitl_shape: "F.8"
+hitl_why: "Opportunities are surfaced across multiple ADD signals as an independent any-order set."
+# Phase 267.3-07, ruled in 267.3-CLASSIFICATION.md (Row 4, navigator-ruled): first delivery at commands/opportunities.md:75, the scan subcommand's live Grants.gov/Simpler Grants search matched against the room's own domain, geography and stage.
+interactive_first_reward: methodology_reframe
+serves_jtbd: ["explore"]
+teaching: "When you need to manage grants as a live pipeline rather than a list, /mos:opportunities surfaces the Opportunity Bank with HSI scores and decision states. Funding deserves the same rigor as product."
+allowed-tools: Read Write Bash Glob Agent WebSearch AskUserQuestion
+# --- Phase 144.1 connector frontmatter ---
+connector:
+  connects_to_spine: true
+  sensor_triggers: []
+  reach_id: context_block
+  sub_mode: opportunities
+  framework: null
+  posture: push_forward
+  hierarchy_rank: 13
+  filing: none
+  plan_gated: false
+  web_scope: null
+---
+
+<!-- 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:opportunities -- Grant Discovery + Opportunity Management
+
+> Context-driven grant discovery. Larry reads your room, generates search queries, and presents opportunities for you to confirm or reject.
+
+## How Opportunities Get Banked
+
+Opportunities enter the bank through two paths:
+
+1. **Grant discovery** (`/mos:opportunities scan`) - external funding opportunities from Grants.gov and Simpler Grants, filed after user confirmation
+2. **Intelligence cascade** (automatic) - every methodology command that triggers analyze-room extracts gaps, convergences, and contradictions as bankable opportunities. These are filed automatically with no user action needed.
+
+Every opportunity carries a Knight position (risk vs uncertainty vs mixed) and a confidence score. Risk = known problem with quantifiable odds. Uncertainty = unknown problem requiring exploration. Mixed = contradiction that could go either way.
+
+## Subcommands
+
+### scan
+
+Run a context-driven grant discovery scan.
+
+**How it works:**
+1. Larry reads your room's STATE.md (domain, geography, venture stage) and problem-definition/ for context
+2. Generates search queries from YOUR room data (not hardcoded searches)
+3. Searches Grants.gov and Simpler Grants APIs concurrently
+4. Deduplicates and scores results by relevance to your room
+5. Presents results in a table with relevance reasoning
+
+**After presenting results, ask the user:**
+
+> I found {N} opportunities relevant to your venture. Here are the top matches:
+>
+> | # | Funder | Program | Amount | Deadline | Relevance |
+> |---|--------|---------|--------|----------|-----------|
+> | 1 | ... | ... | ... | ... | 0.85 |
+>
+> **What would you like to do?**
+> - **File all** -- I'll file all opportunities to your opportunity-bank
+> - **Review individually** -- I'll walk through each one for your decision
+> - **Skip** -- No opportunities filed
+
+This is the **confirm-first pattern**: Larry presents, user decides. Nothing is filed automatically.
+
+**CLI:** `mindrian-tools.cjs opportunity scan [roomDir]`
+**MCP:** `data_room` tool with command `scan-opportunities`
+
+### list
+
+Show filed opportunities with optional filters (--domain, --knight, --min-confidence).
+
+**Display format:**
+
+> | # | Source | Program/Problem | Knight Position | Confidence | Deadline | Status |
+> |---|--------|-----------------|-----------------|------------|----------|--------|
+> | 1 | NIH | SBIR Phase I | risk | 0.85 | 2026-06-01 | filed |
+> | 2 | diagnose | Missing structural coverage in market-analysis | uncertainty | 0.8 | - | banked |
+
+This unified table works for both grant-scanned and cascade-extracted opportunities:
+- Grant opportunities: Source = funder name, Knight Position = 'risk' (grants are known risk), Confidence = relevance_score
+- Cascade opportunities: Source = source_framework, Knight Position = from schema, Confidence = from schema
+
+### Filter Flags
+
+The list subcommand supports filter flags to narrow results:
+
+- `--domain <name>` -- Filter by domain (partial match, case-insensitive). Example: `--domain market-analysis`
+- `--knight <position>` -- Filter by Knight position: `risk`, `uncertainty`, or `mixed`
+- `--min-confidence <N>` -- Filter by minimum confidence score (0.0-1.0). Example: `--min-confidence 0.7`
+
+Flags can be combined: `--domain healthcare --knight uncertainty --min-confidence 0.5`
+
+**CLI:** `mindrian-tools.cjs opportunity list [roomDir] --domain X --knight Y --min-confidence N`
+**MCP:** `data_room` tool with command `list-opportunities`
+
+### file
+
+File a specific opportunity after scan confirmation. Used internally after user confirms from scan results.
+
+**CLI:** `mindrian-tools.cjs opportunity file [roomDir] [dataJson]`
+**MCP:** `data_room` tool with command `file-opportunity`
+
+### bank
+
+List all banked opportunities extracted from methodology commands (not grant scans).
+
+These are opportunities discovered by the intelligence cascade - gaps, convergences, and contradictions that represent bankable problems.
+
+**Display format:**
+
+> Here are the opportunities your Data Room intelligence has banked:
+>
+> | # | Problem | Domain | Knight Position | Confidence | Source | Status |
+> |---|---------|--------|-----------------|------------|--------|--------|
+> | 1 | Missing structural coverage in market-analysis | market-analysis | uncertainty | 0.8 | diagnose | banked |
+>
+> **Actions:**
+> - **View details** - I'll show the full evidence and mirror solution
+> - **Pursue** - promote to funding pipeline
+> - **Reject** - capture your reason (this teaches me)
+
+**CLI:** `mindrian-tools.cjs opportunity bank [roomDir]`
+**MCP:** `data_room` tool with command `bank-opportunities`
+
+## Rejection Handling
+
+When the user rejects an opportunity, **capture the reason**. Rejection is data (per CLAUDE.md architecture).
+
+Ask: "Why are you passing on this one? (This helps me find better matches next time.)"
+
+The reason is recorded in opportunity-bank/STATE.md and informs future scans.
+
+## Insufficient Context
+
+If the room lacks sufficient context for grant discovery (no domain_keywords, sparse problem-definition), Larry should explain:
+
+> "I need more context about your venture to search for relevant grants. Your room needs:
+> - `domain_keywords` in STATE.md (e.g., artificial-intelligence, healthcare)
+> - Content in problem-definition/ describing your domain
+> - `geography` in STATE.md for eligibility matching
+>
+> Run /mos:room update to add this context, then try scanning again."
+
+## Discovery is Context-Driven
+
+The key insight: Larry does NOT search for "AI grants" because someone hardcoded that. Larry searches for grants that match THIS room's specific problem domain, geography, team profile, and venture stage. Every room gets different search queries.
+
+The room context generates the search queries:
+- `domain_keywords` --> API keyword parameters and funding categories
+- `geography` --> eligibility filters
+- `venture_stage` --> grant type matching (SBIR for pre-revenue, etc.)
+- `problem-definition/` content --> additional domain terms

package/skills/organize/SKILL.md

@@ -0,0 +1,69 @@
+---
+name: organize
+description: "[Deprecated] Navigate room hierarchy with graph-aware tree (use /mos:rooms organize)"
+license: BSL-1.1. See LICENSE for complete terms (Business Source License 1.1, Change Date 2030-04-16 to Apache License 2.0).
+help_jtbd: "Group rooms into portfolios (deprecated: use /mos:rooms organize)."
+argument-hint: "[tree|propose|move <room> <group>]"
+body_shape: B (Semantic Tree)
+hitl_shape: "F.8"
+hitl_why: "Filing and tidy jobs are surfaced as an independent set with no ordering constraint."
+# Phase 267.3-07, ruled in 267.3-CLASSIFICATION.md (Row 5): first delivery at commands/organize.md:58, a soft-alias stub passing through /mos:rooms organize's output verbatim (router sub-case).
+interactive_first_reward: "--none (diagnostic surface)"
+serves_jtbd: ["audit-room"]
+deprecated: true
+deprecated_redirect: "rooms organize"
+deprecated_removal: "v1.14.0"
+teaching: "Deprecated alias. Use /mos:rooms organize to manage the room portfolio hierarchy; organize is folded into the multi-room surface. Scheduled removal: v1.14.0."
+ui_reference: skills/ui-system/SKILL.md
+allowed-tools: Read Write Bash Glob Grep AskUserQuestion
+# --- Phase 172-06 CIRS R1 exclude (Canon Part 11) ---
+connector:
+  excluded: true
+  reason: "Utility command. Reorganizes / files room artifacts on explicit navigator request; a housekeeping action today. INV-06 promotion candidate (a future contextual trigger on filing pressure is plausible), excluded for now."
+---
+
+<!-- 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:organize
+
+> Deprecated. /mos:organize now redirects to /mos:rooms organize. Scheduled removal: v1.14.0. Use /mos:rooms organize going forward.
+
+You are Larry. The user invoked /mos:organize. Per D-09 (LOCKED 2026-05-16, Phase 121.5-08 Sub-plan J) /mos:organize is a soft-alias stub for the v1.13.x window. The canonical surface is /mos:rooms organize.
+
+## Steps
+
+1. Emit the deprecation note above as a single cyan line (Larry voice; no em-dash; one sentence per skills/ui-system/SKILL.md Section 6).
+
+2. Invoke /mos:rooms organize with the user's original arguments (the subcommand verb: `tree`, `propose`, `view`, `move`). Run:
+
+```bash
+node "${MINDRIAN_OS_ROOT:-${CLAUDE_PLUGIN_ROOT:?MindrianOS install root not found. Set MINDRIAN_OS_ROOT (see lib/core/active-plugin-root.cjs) or run from Claude Code.}}/scripts/soft-alias-runner.cjs" --from organize --to "rooms organize" --remaining-args $ARGUMENTS
+```
+
+The runner emits `{redirect, deprecation_note, args, ok}`. Use the redirect to confirm the target, then proceed with /mos:rooms organize behavior. The user sees ONE deprecation note + the rooms organize output (Shape B Semantic Tree).
+
+3. Pass through /mos:rooms organize's output (the room portfolio tree, the propose listing, or the move confirmation) verbatim.
+
+## Why this is a soft-alias
+
+Cluster 5 audit (2026-05-15) found that /mos:organize and the multi-room /mos:rooms surface both managed the room portfolio hierarchy. Two entry points, one feature. Folding organize into /mos:rooms organize gives one canonical surface for portfolio management.
+
+Per D-09 the old command stays as a soft-alias stub for v1.13.x; removal is scheduled v1.14.0. CHANGELOG announces the rename.
+
+## Cross-references
+
+- `commands/rooms.md` -- the canonical target with the full portfolio surface (and the `organize` subcommand).
+- `scripts/soft-alias-runner.cjs` -- the shared runner.
+- Canon Part 7 (Reuse Before Build) -- consolidation rationale.