@receiz/ai-skills 126.0.0 → 127.0.0

package/README.md

@@ -1,6 +1,9 @@
 # Receiz AI Skills
 
-The `126.0.0` package at release coordinate `v126.0.0` is a registry-bound operating-contract distribution beneath Receiz proof authority. AI instructions, MCP confirmations, OAuth/OIDC grants, JSON objects, database rows, staging references, and receipts never become proof or runtime authority.
+Explore the [public capability directory](https://receiz.com/capabilities) for user outcomes, every public SDK callable, MCP tools, and AI skills. Agents can read the [machine-readable catalog](https://receiz.com/capabilities.json). An inventory entry grants no access and does not imply an MCP execution adapter.
+
+
+The `127.0.0` package at release coordinate `v127.0.0` is a registry-bound operating-contract distribution beneath Receiz proof authority. AI instructions, MCP confirmations, OAuth/OIDC grants, JSON objects, database rows, staging references, and receipts never become proof or runtime authority.
 
 KKSv1.0 is the single temporal authority for proof. New proof coordinates are computed locally from Genesis and the canonical pulse frequency; existing coordinates are carried by the verified sealed object. `createdAt` is descriptive Chronos only and is never proof, ordering, or causal authority. The SDK, MCP tools, all 37 machine manifests, and the shared [KKSv1.0 temporal-authority contract](resources/kks-v1-temporal-authority.md) enforce the same boundary.
 
@@ -12,19 +15,23 @@ Every AI skill is the reader layer beneath those authorities. The manifests
 forbid `reader-as-authority`; an agent may inspect, verify, and project, but it
 may not author, reorder, replace, or outrank the actual sealed proof object.
 
-## V126 — Complete Source and Release Continuity
+## V127 — Reality-Grade Infrastructure
+
+V127 aligns the SDK, compiler, MCP and AI packages with complete source custody, first-seal ownership, deterministic coordinates and verified append continuity. The public function catalog covers 678 callable exports across the main, compiler, testing, React, offline sealing, local subject host and offline KaiSigil entry points. MCP capability reporting includes this catalog and the complete 425-method client inventory; 330 explicit tools retain their SDK authority boundaries.
+
+V127 agents can create full KaiSigil proofs locally with packaged resources and use the [source-bound local subject runtime](../packages/receiz-sdk/docs/local-subject-runtime.md). Temporal execution admits complete causal coordinates; bare pulse values, model reports and copied plans cannot advance authority. [Offline file sealing](../packages/receiz-sdk/docs/offline-sealing.md) runs without network after explicit signer enrollment; KaiSigil proving requires no enrollment or network. Portable history retains exact V120 snapshots and full temporal evidence under the enclosing sealed source.
 
-V126 aligns the SDK, compiler, MCP and AI packages with complete source custody, first-seal ownership, deterministic coordinates and verified append continuity. The public function catalog covers 663 callable exports across the main, compiler, testing and React entry points. MCP capability reporting includes this catalog and the complete 421-method client inventory; 221 explicit tools retain their SDK authority boundaries.
+Capability inventory is not a claim that every historical HTTP route exists. Of the 81 newly mapped JSON SDK adapters, 12 compose current routes and 69 require explicit historical HTTP compatibility; older `media.upload`, `media.transform`, and `proof.query` also require that transport. Default calls fail before requests or mutation previews. Agents must follow [historical HTTP compatibility](../packages/receiz-sdk/docs/historical-http.md) and the separate historical subject transport boundary in the local runtime guide. Neither compatibility configuration nor credentials become proof authority.
 
 ### Preserved V125 economy contracts
 
 **Sealed Reality Becomes Spendable Value. Economic authority follows proof, not institution.** V125 teaches agents to register a sealed lawful-action law, submit a separately sealed completed action whose causal claims are derived from a reverified portable execution-authority transition, and request deterministic current-Kai Settlement projection without inventing value. The exact micro-Phi law is `(completionKai - startKai + 1) * (valuationKai - completionKai + 1)`. No AI, developer, government, institution, server, database, or model may supply a completion verdict, amount, multiplier, bonus, backfill, or proof exemption.
 
-The current 221-tool MCP registry preserves the complete 169-tool v124.1 runtime inventory and adds 39 trust adapters and thirteen v125 economy adapters: three lawful-action operations plus the complete ten-operation sender/receiver edge lifecycle for planning, inspection, transition-set verification, commit preparation, recovery packaging, Settlement send/receive, and Reserve send/receive. Exact proof, plan, transition, commit, and recovery materials stay in trusted-host or same-runtime custody behind opaque references. The enclosing sealed artifact is verified before payload extraction. One admitted membership belongs to the current owner; transfer moves it without cloning Settlement value or erasing provenance. Reserve remains the legacy funding rail.
+The current 330-tool MCP registry preserves the 169-tool v124.1 runtime inventory, 39 trust adapters, and thirteen v125 economy adapters alongside the new v127 tools. The thirteen economy adapters comprise three lawful-action operations plus the complete ten-operation sender/receiver edge lifecycle for planning, inspection, transition-set verification, commit preparation, recovery packaging, Settlement send/receive, and Reserve send/receive. Exact proof, plan, transition, commit, and recovery materials stay in trusted-host or same-runtime custody behind opaque references. The enclosing sealed artifact is verified before payload extraction. One admitted membership belongs to the current owner; transfer moves it without cloning Settlement value or erasing provenance. Reserve remains the legacy funding rail.
 
-`skills.json` also binds all 421 callable methods on the public `createReceizClient()` surface by exact ordered inventory and digest. The union of the 37 machine manifests must contain every method and all 221 MCP tools; generation and validation fail if either distribution surface drifts. Three retained artifact-recovery methods are explicitly classified under `historicalSdkOperations`, remain callable, and are not mislabeled as v125 defaults. `receiz_capabilities` exposes the same SDK inventory, exact MCP tool inventory, and v125 capability families without turning capability description into proof or transaction authority.
+`skills.json` also binds all 425 callable methods on the public `createReceizClient()` surface by exact ordered inventory and digest. The union of the 37 machine manifests must contain every method and all 330 MCP tools; generation and validation fail if either distribution surface drifts. Three retained artifact-recovery methods are explicitly classified under `historicalSdkOperations`, remain callable, and are not mislabeled as v125 defaults. `receiz_capabilities` exposes the same SDK inventory, exact MCP tool inventory, and v125 capability families without turning capability description into proof or transaction authority.
 
-All 37 current manifests use package/ruleset `126.0.0`, SDK/MCP range `>=126.0.0 <127.0.0`, registry digest `80137c2e6f294050ef36ff75e4daac15c7790b7f04d9a91fab9d1970fa3c0b09`, and operation-matrix digest `42c7f0924df91b4ba11c1b891fee2b92abb509430a86b030735c23d055e67949`.
+All 37 current manifests use package/ruleset `127.0.0`, SDK/MCP range `>=127.0.0 <128.0.0`, registry digest `8d0b5b839d02d9efbd4306cc99410595a183705c2670b76d2567eaaaade99065`, and operation-matrix digest `eadd171a45fcc51e275a1c57de1eb8e67614757a5723d141793641edf7207a10`.
 
 ## Historical V124 — Reality Becomes Infrastructure
 
@@ -63,11 +70,11 @@ performance, observability, testing, and release contracts. The earlier nine
 domain skill names remain packaged for compatibility. Seven operation skill names first introduced in obsolete-versioned now
 describe current outcomes for identity profiles, portable continuity, bearer ownership, offline transport,
 proof media, cross-app state, and admission evidence. Their retired obsolete-versioned mechanisms are not active defaults.
-One dedicated `receiz-global-reconciliation` skill governs accepted-head resolution, verified offline reconciliation, structural divergence, first paint, indeterminate recovery, and effects. `receiz-value-execution` governs exact live Phi execution and retry recovery; `receiz-proof-authority` governs edge verification, explicit consent, and minimum-scope embedded authority. The v124 coordinates remain immutable historical evidence; the v125 manifest coordinates above are current.
+One dedicated `receiz-global-reconciliation` skill governs accepted-head resolution, verified offline reconciliation, structural divergence, first paint, indeterminate recovery, and effects. `receiz-value-execution` governs exact live Phi execution and retry recovery; `receiz-proof-authority` governs edge verification, explicit consent, and minimum-scope embedded authority. The v124 coordinates remain immutable historical evidence; the v127 manifest coordinates above are current.
 
 This package teaches AI agents how to understand, verify, build with, and operate Receiz from the existing repository surfaces.
 
-It is published as `@receiz/ai-skills` and is an exact-version dependency of both `@receiz/sdk` and `@receiz/mcp-server`. Installing either package places all 42 skills under `node_modules/@receiz/ai-skills`: 36 manifests and 33 OpenAI agent prompts are included for explicit loading by an agent host.
+It is published as `@receiz/ai-skills` and is an exact-version dependency of both `@receiz/sdk` and `@receiz/mcp-server`. Installing either package places all 43 skills under `node_modules/@receiz/ai-skills`: 37 manifests and 34 OpenAI agent prompts are included for explicit loading by an agent host.
 
 Receiz is a proof-native artifact system. Sealed proof/object truth is strongest: a Receiz object is not a database row, but a proof-carrying artifact whose witnessed history is the truth boundary. The proof object carries continuity; SDK and MCP may verify, extract, append, or project beneath that carried truth. Receiz.com reference behavior comes before SDK, MCP, AI, and other developer rails. Server state, database state, UI state, marketplace state, and model memory are projections only. Old witnessed truth is not stale. If a projection is wrong, rebuild it from object history. If truth is missing, append new truth. Never mutate witnessed truth.
 
@@ -76,7 +83,7 @@ A verified proof object is not limited to the platform that created it. Any lawf
 ## SDK, MCP, And Skills
 
 - SDK: `@receiz/sdk` is the universal runtime; `@receiz/sdk/react` is React-only; `@receiz/sdk/compiler` is Node-only; `@receiz/sdk/testing` is browser-safe sandbox and conformance support.
-- MCP: `@receiz/mcp-server` is for agent tool access. Its current 221-tool registry includes the complete 169-tool v124.1 inventory plus thirteen v125 lawful-action and edge-value tools and 39 trust adapters. The four V124.1 source-carried adapters remain unchanged and expose replay-family opening, Receiz-ID conversation-family opening, bounded private epoch-grant planning, and held material-family inspection while exact sources, secrets, locators, and large bytes remain in trusted-host custody.
+- MCP: `@receiz/mcp-server` is for agent tool access. Its current 330-tool registry preserves the 169-tool v124.1 inventory, thirteen v125 lawful-action and edge-value tools, and 39 trust adapters alongside the v127 tools. The four V124.1 source-carried adapters remain unchanged and expose replay-family opening, Receiz-ID conversation-family opening, bounded private epoch-grant planning, and held material-family inspection while exact sources, secrets, locators, and large bytes remain in trusted-host custody.
 
 `skills.json` carries the canonical `v124_1CapabilityInventory`, and each relevant manifest carries its exact `v124_1CapabilityIds`. This binds source-carried replay, unlimited invited-member conversation, exact large-material reconstruction, progressive verified playback, proof-derived memory/Twin response, and portable account continuity across SDK, MCP, and AI instructions without allowing the model layer to invent or outrank proof.
 - Skills: this package is operating doctrine. It tells an AI which Receiz primitive is active, which source of truth wins, which SDK or MCP rail to use, and what must never be assumed.
@@ -85,7 +92,7 @@ Agents may acquire scoped delegated access through the official Receiz Connect/O
 
 ## How Agents Should Use This Package
 
-Start with `receiz-build-production-system` when work spans multiple constitutional domains and `receiz-global-reconciliation` for named-domain coordination. Machine-readable manifests require the exact active canonical v126 registry and operation-matrix digests, command-only mutation, independent verification, MCP conformance, and a passing release lock. Use a focused operation skill for SDK/MCP implementation and a constitutional skill for broader system law. Stable obsolete-versioned wire schemas remain available only through the explicit historical `@receiz/sdk/obsolete-versioned` package entry; current skills never teach their key, head, receipt, reconcile, or signed-command mechanics as a current outcome.
+Start with `receiz-build-production-system` when work spans multiple constitutional domains and `receiz-global-reconciliation` for named-domain coordination. Machine-readable manifests require the exact active canonical v127 registry and operation-matrix digests, command-only mutation, independent verification, MCP conformance, and a passing release lock. Use a focused operation skill for SDK/MCP implementation and a constitutional skill for broader system law. Stable obsolete-versioned wire schemas remain available only through the explicit historical `@receiz/sdk/obsolete-versioned` package entry; current skills never teach their key, head, receipt, reconcile, or signed-command mechanics as a current outcome.
 
 The current operation skills are executable contracts. Current profile work uses the neutral authenticated same-UID profile operation. Current bearer ownership accepts a complete verified artifact, derives prior ownership from carried proof, and returns a native Record -> Seal artifact. Proof media uses the same native artifact custody and projects only a verified URL beneath it. These outcomes have no active identity-key, caller-head, claim-key, or receipt prerequisite. MCP calls the same SDK outcomes and never creates a parallel authority.
 
@@ -128,6 +135,6 @@ When SDK or MCP changes:
 - `receiz-event-derived-memory` governs factual memories with admitted event citations and rebuildable AI summaries/projections.
 - `receiz-live-proof-character` governs streamed reply, audio, viseme, gaze, blink, breath, emotion, gesture, proposed intention, and completion events as non-authoritative performance over exact proof context.
 
-Every living-subject skill carries its SDK map, MCP map, authority rules, examples, tests, manifest, OpenAI metadata, and the exact current v126 registry and operation-matrix digests. The skills teach the same sequence as the implementation: canonical head → entire proof history → compact indexes → relevant references → exact primary proof objects → reasoning → cited answer. They never authorize a world event, transfer, mandate, or release claim by prose.
+Every living-subject skill carries its SDK map, MCP map, authority rules, examples, tests, manifest, OpenAI metadata, and the exact current v127 registry and operation-matrix digests. The skills teach the same sequence as the implementation: canonical head → entire proof history → compact indexes → relevant references → exact primary proof objects → reasoning → cited answer. They never authorize a world event, transfer, mandate, or release claim by prose.
 
 The updated constitutional skills bind living-subject behavior into app building, command construction, deterministic replay, causal sync, global reconciliation, portable continuity, authority/security, MCP operation, testing, release, constitutional law, and offline-first operation. Their examples cover exact proof-memory speech, absent-owner exploration, mutual relationship formation, bounded trade, atomic battle, cross-device restoration, mandate revocation during a queued action, partition convergence, rejection of an AI-invented event, and cross-application subject continuity.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@receiz/ai-skills",
-  "version": "126.0.0",
+  "version": "127.0.0",
   "description": "Registry-bound Receiz AI operating contracts beneath proof authority for SDK and MCP agents.",
   "type": "module",
   "files": [