minutework 0.1.44 → 0.1.46

package/assets/claude-local/CLAUDE.md.template

@@ -42,6 +42,12 @@ workspace state. It is wired for Cursor (`.cursor/mcp.json`) and Codex
 `mcp/claude-desktop.sample.json`. Refresh all of this with
 `minutework workspace sync-assets`.
 
+Primitive catalog membership does not imply executability or agent exposure;
+check the live-status block in skills/runtime-capability-inventory/SKILL.md and
+the minutework_runtime_primitive_status workspace MCP tool before authoring
+against a primitive (interim paths:
+skills/runtime-primitive-interim-paths/SKILL.md).
+
 ## Hosted Sandbox Runtime
 
 Use `minutework sandbox create` to provision and link the workspace to its

package/assets/claude-local/skills/README.md

@@ -52,6 +52,7 @@ Generated-workspace-first guidance should live here, especially:
 - `workspace-guidance-refresh/SKILL.md`
 - `shell-architecture/SKILL.md`
 - `runtime-capability-inventory/SKILL.md`
+- `runtime-primitive-interim-paths/SKILL.md`
 - `integration-broker-and-connectors/SKILL.md`
 - `layering-and-import-modes/SKILL.md`
 - `standalone-mobile-client/SKILL.md`

package/assets/claude-local/skills/ai-capability-defaults/SKILL.md

@@ -23,6 +23,8 @@ generation, itinerary generation, content generation, or structured-output UX.
 - Start with `minutework_capability_inventory` and the generated workspace
   capability guidance before assuming runtime AI capability is present,
   missing, or unsupported.
+- When a needed primitive is catalogued but not yet executable or agent-exposed,
+  use the sanctioned interim paths in `runtime-primitive-interim-paths/SKILL.md`.
 - If runtime AI availability is unclear, inspect the linked runtime/workspace
   context or ask a clarifying question before inventing new AI infrastructure.
 - When AI drafting or generation touches external participants, guests,

package/assets/claude-local/skills/published-web-and-mw-core-site/SKILL.md

@@ -77,11 +77,11 @@ flows, or the default MinuteWork site model.
 ## Current Baseline Catalog
 
 <!-- builder-doc-sync:mw-core-site:start -->
-- Baseline schemas: `mw.core.site.config`, `mw.core.site.page`,
-  `mw.core.site.nav`, `mw.core.site.form`, `mw.core.site.submission`
+- Baseline schemas: `mw.core.site.config`, `mw.core.site.page`, `mw.core.site.nav`,
+  `mw.core.site.form`, `mw.core.site.submission`
 - Query surfaces: `site.get_public_config`, `site.get_page_by_path`,
   `site.list_navigation`, `site.get_form_definition`
-- Action surfaces: `site.upsert_config`, `site.upsert_page`,
-  `site.upsert_navigation`, `site.upsert_form`
+- Action surfaces: `site.upsert_config`, `site.upsert_page`, `site.upsert_navigation`,
+  `site.upsert_form`
 - Reserved next schema: `mw.core.site.post`
 <!-- builder-doc-sync:mw-core-site:end -->

package/assets/claude-local/skills/runtime-capability-inventory/SKILL.md

@@ -92,19 +92,70 @@ generated Builder workspace.
 ## Known Seeded Primitive Catalog
 
 <!-- builder-doc-sync:active-primitives:start -->
-- Current active primitives by family:
-  - `records`: `records.query`, `records.upsert`
-  - `workflow`: `workflow.run_action`, `workflow.schedule_timer`,
-    `workflow.resume_checkpoint`, `workflow.cancel_on_thread_activity`
-  - `projection`: `projection.publish_receipt`
-  - `integrations`: `integrations.query_provider`,
-    `integrations.execute_automation`
-  - `communication`: `communication.send_email`, `communication.send_sms`
-  - `consent`: `consent.check_channel`
-  - `scheduling`: `scheduling.book_appointment`
-  - `discovery`: `discovery.request_proposal`
-  - `voice`: `voice.initiate_call`, `voice.hangup_call`,
-    `voice.get_call_status`
-  - `ontology`: `ontology.resolve_local`, `ontology.propose_candidate`
-  - `command_exec`: `command_exec.run_template`
+- Three-layer rule: catalogued != executor-wired != agent-exposed. Catalog membership
+  never implies a primitive can execute today, and an executor never implies any agent
+  can call it; check each layer separately.
+- Default agent ring is the read-only READ_RING (`records.query`,
+  `app_pack.list_surfaces`, `app_pack.query`); its members need no grant. (`app_pack.*`
+  and `delegate_to_builder` are agent tools, not catalog primitives.) Every other tool
+  needs a per-agent `enable_agent_tools` grant (operator-only). Enabling the effectful
+  `app_pack.action` tool is additionally process-gated on the A1-A3 human-approval
+  surface being live for the tenant plane; surfaces with forced approval park on a human
+  approval before executing.
+- Machine-readable copy with per-primitive live-status fields:
+  `skills/runtime-capability-inventory/primitive-catalog.json` (synced from the runtime
+  fixture). The `minutework_runtime_primitive_status` workspace MCP tool serves the same
+  answer, linked-runtime-backed when a runtime binding exists.
+- Live executors (catalogued and wired today):
+  - `records.query` -- Native agent tool in the default READ_RING: enabled for the
+    provisioning-default agent without any grant.
+  - `records.upsert` -- Effectful; outside the default READ_RING, granted per agent via
+    enable_agent_tools (OPERATOR_RING member).
+  - `workflow.run_action` -- Agents reach this envelope through the app_pack.action tool
+    over manifest-declared agent_exposed action surfaces; approval is decided per target
+    capability/definition.
+  - `workflow.schedule_timer` -- Workflow-internal machinery (WORKFLOW_TIMER jobs); not
+    an agent-callable tool.
+  - `workflow.resume_checkpoint` -- Workflow-internal machinery driven by the jobs
+    worker; not an agent-callable tool.
+  - `workflow.cancel_on_thread_activity` -- Workflow-internal machinery; not an
+    agent-callable tool.
+  - `integrations.query_provider` -- Read-only provider query; preflight fail-closes on
+    connection/secret availability.
+  - `integrations.execute_automation` -- Approval is definition-forced in
+    mw.core.agent_actions (a definition can force approval, never relax it).
+  - `consent.check_channel` -- Service seam consumed by other executors and sidecars as
+    a preflight; not an agent-callable tool.
+  - `discovery.request_proposal` -- Effectful; outside the default READ_RING, granted
+    per agent via enable_agent_tools (BUILDER_DELEGATION_RING member).
+  - `command_exec.run_template` -- Operator/platform-dispatch surface; not an
+    agent-callable tool.
+  - `voice.hangup_call` -- VoiceSession finalization is live; the provider-side hangup
+    leg is platform-owned Phase 4. Not an agent-callable tool.
+  - `voice.get_call_status` -- Read-only session status lookup. Not an agent-callable
+    tool.
+- Partial executors (wired with a deferred leg; treat the deferred
+  leg as unavailable):
+  - `scheduling.book_appointment` -- Capability is registered but the direct booking
+    executor is MVP-deferred; preflight truthfully reports provider_unavailable.
+  - `voice.initiate_call` -- Pre-dial stack is live (consent + TCPA preflight, approval
+    parking, VoiceSession lifecycle); the Twilio dial + realtime media worker are
+    platform-owned Phase 4.
+- Catalogued-only (no executor wired; use the sanctioned interim
+  path):
+  - `projection.publish_receipt` -- Owning services emit receipts through the runtime
+    bridge directly (e.g. command_exec posts its own run receipts); do not build a
+    bespoke generic receipt executor.
+  - `communication.send_email` -- Sidecar bridge mirroring the
+    runtime_bridge/integrations transport seam with a consent.check_channel preflight;
+    route sends through the send_notification seam, never inline in runtime. End-state
+    is an action capability like voice.initiate_call.
+  - `communication.send_sms` -- Sidecar bridge mirroring the runtime_bridge/integrations
+    transport seam with a consent.check_channel preflight; route sends through the
+    send_notification seam, never inline in runtime. End-state is an action capability
+    like voice.initiate_call.
+  - `ontology.resolve_local` -- Seed URNAnchor rows at install; treat resolution as a
+    read-only lookup over seeded anchors.
+  - `ontology.propose_candidate` -- Stage PromotionCandidate rows for manual review; the
+    promotion service is platform-owned, do not build it tenant-side.
 <!-- builder-doc-sync:active-primitives:end -->

package/assets/claude-local/skills/runtime-capability-inventory/primitive-catalog.json

@@ -0,0 +1,392 @@
+{
+  "catalog_version": "1.1.0",
+  "live_status_contract": {
+    "note": "Live-status fields are hand-maintained claims validated against the in-process registries by apps/runtime_builder/tests/test_primitive_live_status.py (fail-closed: a claim without registry evidence fails CI). Catalog membership, executor wiring, and agent exposure are three separate layers.",
+    "executor_ready_values": ["live", "partial", "none"],
+    "agent_exposure_path_values": ["native_tool", "app_pack_action", "none"],
+    "approval_required_null_means": "no registry-evidenced approval seam exists for this primitive: either no executor is wired, or approval is decided per target capability/definition (the workflow.run_action envelope)",
+    "executor_ref_schemes": ["tool:<registered agent tool name>", "capability:<registered action capability id>", "service:<module>:<attr>"]
+  },
+  "primitives": [
+    {
+      "primitive_id": "records.query",
+      "family": "records",
+      "display_name": "Query Records",
+      "description": "Query dynamic entities by schema, filters, and paging.",
+      "runtime_app": "runtime_app_host",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "tool:records.query",
+      "executor_notes": "Native agent tool in the default READ_RING: enabled for the provisioning-default agent without any grant.",
+      "agent_exposable": true,
+      "agent_exposure_path": "native_tool",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "records.upsert",
+      "family": "records",
+      "display_name": "Upsert Record",
+      "description": "Create or update a dynamic entity against a registered schema.",
+      "runtime_app": "runtime_app_host",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "tool:records.upsert",
+      "executor_notes": "Effectful; outside the default READ_RING, granted per agent via enable_agent_tools (OPERATOR_RING member).",
+      "agent_exposable": true,
+      "agent_exposure_path": "native_tool",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "workflow.run_action",
+      "family": "workflow",
+      "display_name": "Run Action",
+      "description": "Execute a typed action through the action-execution contract.",
+      "runtime_app": "runtime_app_host",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": true,
+      "executor_ready": "live",
+      "executor_ref": "service:apps.runtime_app_host.services:execute_action_run",
+      "executor_notes": "Agents reach this envelope through the app_pack.action tool over manifest-declared agent_exposed action surfaces; approval is decided per target capability/definition.",
+      "agent_exposable": true,
+      "agent_exposure_path": "app_pack_action",
+      "approval_required": null,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "workflow.schedule_timer",
+      "family": "workflow",
+      "display_name": "Schedule Timer",
+      "description": "Schedule a durable timer, reminder, or milestone.",
+      "runtime_app": "runtime_jobs",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "service:apps.runtime_app_host.services:schedule_workflow_timer",
+      "executor_notes": "Workflow-internal machinery (WORKFLOW_TIMER jobs); not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "workflow.resume_checkpoint",
+      "family": "workflow",
+      "display_name": "Resume Checkpoint",
+      "description": "Resume a paused workflow at a human-approval or wait checkpoint.",
+      "runtime_app": "runtime_app_host",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "service:apps.runtime_app_host.services:resume_workflow_checkpoint",
+      "executor_notes": "Workflow-internal machinery driven by the jobs worker; not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "workflow.cancel_on_thread_activity",
+      "family": "workflow",
+      "display_name": "Cancel On Thread Activity",
+      "description": "Arm a workflow run to cancel when its thread sees inbound activity (e.g. a reply or unsubscribe).",
+      "runtime_app": "runtime_app_host",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "service:apps.runtime_app_host.services:arm_workflow_cancel_on_thread_activity",
+      "executor_notes": "Workflow-internal machinery; not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "projection.publish_receipt",
+      "family": "projection",
+      "display_name": "Publish Receipt",
+      "description": "Emit a receipt to Core through the runtime bridge.",
+      "runtime_app": "runtime_bridge",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "none",
+      "executor_ref": "",
+      "executor_notes": "Catalogued only: no generic executor is wired for this primitive id.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": null,
+      "interim_path": "Owning services emit receipts through the runtime bridge directly (e.g. command_exec posts its own run receipts); do not build a bespoke generic receipt executor."
+    },
+    {
+      "primitive_id": "integrations.query_provider",
+      "family": "integrations",
+      "display_name": "Query Provider",
+      "description": "Query an external provider through a managed integration connection.",
+      "runtime_app": "runtime_integrations",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": true,
+      "executor_ready": "live",
+      "executor_ref": "capability:integrations.query_provider",
+      "executor_notes": "Read-only provider query; preflight fail-closes on connection/secret availability.",
+      "agent_exposable": true,
+      "agent_exposure_path": "app_pack_action",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "integrations.execute_automation",
+      "family": "integrations",
+      "display_name": "Execute Automation",
+      "description": "Run a browser-automation or portal-execution job.",
+      "runtime_app": "runtime_integrations",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": true,
+      "executor_ready": "live",
+      "executor_ref": "capability:integrations.execute_automation",
+      "executor_notes": "Approval is definition-forced in mw.core.agent_actions (a definition can force approval, never relax it).",
+      "agent_exposable": true,
+      "agent_exposure_path": "app_pack_action",
+      "approval_required": true,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "communication.send_email",
+      "family": "communication",
+      "display_name": "Send Email",
+      "description": "Send an outbound email through a configured provider.",
+      "runtime_app": "runtime_communication",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": true,
+      "executor_ready": "none",
+      "executor_ref": "",
+      "executor_notes": "Catalogued only: no send executor is registered for this primitive id.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": null,
+      "interim_path": "Sidecar bridge mirroring the runtime_bridge/integrations transport seam with a consent.check_channel preflight; route sends through the send_notification seam, never inline in runtime. End-state is an action capability like voice.initiate_call."
+    },
+    {
+      "primitive_id": "communication.send_sms",
+      "family": "communication",
+      "display_name": "Send SMS",
+      "description": "Send an outbound SMS through a configured provider.",
+      "runtime_app": "runtime_communication",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": true,
+      "executor_ready": "none",
+      "executor_ref": "",
+      "executor_notes": "Catalogued only: no send executor is registered for this primitive id.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": null,
+      "interim_path": "Sidecar bridge mirroring the runtime_bridge/integrations transport seam with a consent.check_channel preflight; route sends through the send_notification seam, never inline in runtime. End-state is an action capability like voice.initiate_call."
+    },
+    {
+      "primitive_id": "consent.check_channel",
+      "family": "consent",
+      "display_name": "Check Channel Consent",
+      "description": "Preflight a recipient/channel for suppression, contact preferences, marketing consent, quiet hours, and rate limits before sending.",
+      "runtime_app": "runtime_communication",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "service:apps.runtime_communication.services:check_channel",
+      "executor_notes": "Service seam consumed by other executors and sidecars as a preflight; not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "scheduling.book_appointment",
+      "family": "scheduling",
+      "display_name": "Book Appointment",
+      "description": "Book a scheduling appointment through a configured provider (Cal.com/Calendly). MVP: the agent shares a correlation-tokened booking link and the provider webhook confirms the booking; the direct executor is provider-gated.",
+      "runtime_app": "runtime_integrations",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": false,
+      "executor_ready": "partial",
+      "executor_ref": "capability:scheduling.book_appointment",
+      "executor_notes": "Capability is registered but the direct booking executor is MVP-deferred; preflight truthfully reports provider_unavailable.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": "Target the workflow.run_action link+webhook envelope: share a correlation-tokened booking link (consent-guarded), and the provider webhook POST /api/v1/ingress/booking confirms the booking. Only the preflight changes when the direct executor lands."
+    },
+    {
+      "primitive_id": "discovery.request_proposal",
+      "family": "discovery",
+      "display_name": "Request Discovery Proposal",
+      "description": "Bridge a thread into the Vuilder discovery engine: create a discovery lead/session (resume optional), link it to the thread, and poll the gateway until the proposal is ready, then post the build-direction cards member-only. Discovery stays authoritative in public-dj.",
+      "runtime_app": "runtime_agents",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": true,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "tool:discovery.request_proposal",
+      "executor_notes": "Effectful; outside the default READ_RING, granted per agent via enable_agent_tools (BUILDER_DELEGATION_RING member).",
+      "agent_exposable": true,
+      "agent_exposure_path": "native_tool",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "ontology.resolve_local",
+      "family": "ontology",
+      "display_name": "Resolve Local Ontology",
+      "description": "Resolve a local record against the ontology bundle and Core URNs.",
+      "runtime_app": "runtime_memory",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "none",
+      "executor_ref": "",
+      "executor_notes": "Catalogued only: URNAnchor models exist but no resolve executor is wired.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": null,
+      "interim_path": "Seed URNAnchor rows at install; treat resolution as a read-only lookup over seeded anchors."
+    },
+    {
+      "primitive_id": "ontology.propose_candidate",
+      "family": "ontology",
+      "display_name": "Propose Ontology Candidate",
+      "description": "Submit a promotion candidate to Core for shared ontology acceptance.",
+      "runtime_app": "runtime_bridge",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "none",
+      "executor_ref": "",
+      "executor_notes": "Catalogued only: PromotionCandidate models exist but no propose executor is wired.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": null,
+      "interim_path": "Stage PromotionCandidate rows for manual review; the promotion service is platform-owned, do not build it tenant-side."
+    },
+    {
+      "primitive_id": "command_exec.run_template",
+      "family": "command_exec",
+      "display_name": "Run Command Template",
+      "description": "Execute a named command template with validated arguments.",
+      "runtime_app": "runtime_command_exec",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": false,
+      "requires_secret": false,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "service:apps.runtime_command_exec.services:launch_command_execution",
+      "executor_notes": "Operator/platform-dispatch surface; not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "voice.initiate_call",
+      "family": "voice",
+      "display_name": "Initiate Call",
+      "description": "Place an outbound voice call to a contact and join an AI on the line. Runs through execute_action_run: preflight enforces consent (do_not_call), TCPA calling windows/quiet hours, and telephony-provider availability, then drives a VoiceSession (direction=outbound). The actual Twilio dial + realtime media worker are platform-owned (v3_voice_replacement Phase 4); outbound calls default to human approval.",
+      "runtime_app": "runtime_agents",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": true,
+      "executor_ready": "partial",
+      "executor_ref": "capability:voice.initiate_call",
+      "executor_notes": "Pre-dial stack is live (consent + TCPA preflight, approval parking, VoiceSession lifecycle); the Twilio dial + realtime media worker are platform-owned Phase 4.",
+      "agent_exposable": true,
+      "agent_exposure_path": "app_pack_action",
+      "approval_required": true,
+      "interim_path": "Build the pre-dial stack now (preflight matrix, APPROVAL_PENDING parking, VoiceSession lifecycle, mock transcripts, hangup/get_call_status); dial, media worker, and inbound telephony are platform-owned Phase 4."
+    },
+    {
+      "primitive_id": "voice.hangup_call",
+      "family": "voice",
+      "display_name": "Hang Up Call",
+      "description": "Finalize an in-progress voice call: finalize the VoiceSession (terminal status + duration) and request provider hangup. Provider hangup is platform-owned (v3_voice_replacement Phase 4).",
+      "runtime_app": "runtime_agents",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "capability:voice.hangup_call",
+      "executor_notes": "VoiceSession finalization is live; the provider-side hangup leg is platform-owned Phase 4. Not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    },
+    {
+      "primitive_id": "voice.get_call_status",
+      "family": "voice",
+      "display_name": "Get Call Status",
+      "description": "Read the current status, duration, and transcript summary for a VoiceSession. Read-only; no provider call.",
+      "runtime_app": "runtime_agents",
+      "dispatch_path": "",
+      "status": "active",
+      "requires_integration": true,
+      "requires_secret": true,
+      "metered": false,
+      "executor_ready": "live",
+      "executor_ref": "capability:voice.get_call_status",
+      "executor_notes": "Read-only session status lookup. Not an agent-callable tool.",
+      "agent_exposable": false,
+      "agent_exposure_path": "none",
+      "approval_required": false,
+      "interim_path": ""
+    }
+  ]
+}

package/assets/claude-local/skills/runtime-primitive-interim-paths/SKILL.md

@@ -0,0 +1,96 @@
+---
+name: runtime-primitive-interim-paths
+description: "Sanctioned interim build paths for catalogued runtime primitives that are not fully executable or agent-exposed yet: outbound comms, booking, ontology, voice, exposure ops, sandbox parity."
+---
+
+# Runtime Primitive Interim Paths
+
+Use this skill when a build targets a runtime primitive that is catalogued but
+not fully executable or agent-exposed yet, and the workspace needs a sanctioned
+interim path instead of bespoke substrate.
+
+Catalog membership never implies executability or agent exposure. Check the
+live-status block in `skills/runtime-capability-inventory/SKILL.md` and the
+`minutework_runtime_primitive_status` workspace MCP tool first; the paths below
+are the sanctioned interim shapes for the known gaps.
+
+## Outbound Email / SMS (`communication.send_email` / `communication.send_sms`)
+
+Status: catalogued, no executor.
+
+- Build outbound delivery as a sidecar bridge mirroring the
+  `runtime_bridge`/integrations transport seam.
+- Always preflight with `consent.check_channel` (suppression, contact
+  preferences, marketing consent, quiet hours, rate limits).
+- Route sends through the runtime's `send_notification` seam, which queues a
+  `SCHEDULED_SEND` job. Never send inline in runtime request paths.
+- The end-state is a governed action capability like `voice.initiate_call`.
+  File a `BuilderCapabilityRequest` gap (see
+  `skills/capability-gap-reporting/SKILL.md`) rather than hard-wiring a
+  provider SDK into runtime.
+
+## Booking (`scheduling.book_appointment`)
+
+Status: partial -- capability registered; preflight truthfully reports
+`provider_unavailable`.
+
+- Target the `workflow.run_action` envelope now: share a correlation-tokened
+  Cal.com/Calendly booking link through a consent-guarded notification.
+- The provider webhook `POST /api/v1/ingress/booking` (this route exists on
+  the platform gateway and the runtime ingress dispatcher) confirms the
+  booking.
+- When the direct executor lands, only the preflight outcome changes. Author
+  flows against the envelope, not the provider.
+
+## Ontology (`ontology.resolve_local` / `ontology.propose_candidate`)
+
+Status: catalogued, no executor; `URNAnchor` and `PromotionCandidate` models
+exist in `runtime_ontology`.
+
+- Seed `URNAnchor` rows at install; treat `resolve_local` as a read-only
+  lookup over seeded anchors.
+- Treat `propose_candidate` as staging `PromotionCandidate` rows for manual
+  review. The promotion service is platform-owned -- do not build it
+  tenant-side.
+
+## Voice (`voice.initiate_call`)
+
+Status: partial; `hangup_call` / `get_call_status` are live.
+
+- The workspace can build the full pre-dial stack: the preflight matrix
+  (consent `do_not_call` -> TCPA calling window 08:00-21:00 callee-local ->
+  provider availability), `APPROVAL_PENDING` parking on the `ActionRun`,
+  `VoiceSession` lifecycle, mock transcripts, and `hangup_call` /
+  `get_call_status` handling.
+- The actual dial, the realtime media worker, and inbound telephony are
+  platform-owned (`v3_voice_replacement` Phase 4). Never implement Twilio
+  internals in a workspace.
+
+## Exposure Ops
+
+- Per-agent tool grants happen only via the operator-run `enable_agent_tools`
+  management command (registry-validated names; `@read` / `@operator` /
+  `@builder_delegation` ring tokens).
+- Manifest-declared exposure (D1) is live: app-pack manifests declare
+  `agentExposed` per action/query (truthy-only, plus approval metadata on
+  capability-bound actions), and those surfaces are reached through the
+  `app_pack.action` / `app_pack.query` tools -- which still require the
+  per-agent grant above.
+- `app_pack.action` exposure is process-gated on the A1-A3 human-approval
+  surface being live for the tenant plane; `integrations.execute_automation`
+  and `voice.initiate_call` park on approval.
+
+## Sandbox Tier Parity
+
+- Sandbox runtimes run identical wiring and gating to primary -- a faithful
+  dry-run. There is no permissive sandbox mode; anything blocked on primary is
+  blocked in sandbox.
+
+## Before Authoring
+
+- Check the live-status block in
+  `skills/runtime-capability-inventory/SKILL.md` and the
+  `minutework_runtime_primitive_status` workspace MCP tool before authoring
+  against any primitive.
+- File gaps via `skills/capability-gap-reporting/SKILL.md` for the comms
+  executor and the ontology services.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "minutework",
-  "version": "0.1.44",
+  "version": "0.1.46",
   "description": "MinuteWork CLI for workspace scaffolding, local preview workflows, and hosted preview deploys.",
   "type": "module",
   "bin": {
@@ -24,8 +24,8 @@
     "@modelcontextprotocol/sdk": "^1.28.0",
     "jiti": "^2.6.1",
     "zod": "^4.3.6",
-    "@minutework/platform-config": "0.1.3",
-    "@minutework/schema-compiler": "0.1.7"
+    "@minutework/schema-compiler": "0.1.8",
+    "@minutework/platform-config": "0.1.3"
   },
   "devDependencies": {
     "@types/node": "^24.9.1",