@llamaventures/cli 2.1.2 → 2.2.0

package/AGENT_BRIEFING.md

@@ -15,9 +15,12 @@ llama agent bootstrap
 The authenticated response contains the current private Llama Brain with
 exactly two components: the approved Investment Framework V3 Markdown (how to
 think) and the Llama Command operating skill (how to work). Beside Brain it
-contains the separate current Live Deal Page field contract (what every visible
-slot means and how to verify a good write). The public CLI intentionally bundles
-none of this private content; it only fetches it after authentication.
+contains a compact index of the current Live Deal Page fields. Before
+`page.patch`, progressively load only the exact fields being changed with
+`llama page-schema read <field> [field...]` or `get_live_deal_page_schema`.
+Load one section only when the write genuinely spans that section. The public
+CLI intentionally bundles none of this private content; it only fetches it
+after authentication.
 
 CLI is the Agent's tool, not Brain. Do not substitute a private skill wrapper
 for either canonical Brain component.
@@ -60,7 +63,7 @@ judgment with its author; otherwise label the value as an Agent proposal / 待
 replace the entire array. Before changing one `whatTheyDo`, `peopleSteps`,
 `foundersJson`, or `industryMaps` item, read the current Page, preserve every
 sibling item, and submit the complete updated array. The live field contract
-includes the exact payload path and write mode for all 61 semantic fields.
+read for those exact fields includes their payload paths and write modes.
 
 ## Exactly four Deal actions
 
@@ -159,16 +162,27 @@ Always submit raw user information first when losing it would be unacceptable:
 }
 ```
 
-Then, when useful, write a structured Information unit:
+Then, when useful, write a structured Information unit. Before each one, ask a
+single question: is this a Llama user's own subjective judgment? Subjective
+means the truth depends on who judges: a value or comparative word with no
+stated benchmark, or a claim about someone's future behavior or inner state. If
+yes, the type is `human_subjective_view.people` (founders or team) or
+`human_subjective_view.business` (company, product, market or deal),
+`value.speaker` names that user, and `value.rawText` quotes the user verbatim
+from `origin.originalUserUtterance`. CLI and Core reject anything else. If no,
+choose any type that fits; founder words are `founder_claim`.
 
 ```json
 {
   "operation": "information.put",
   "dealId": "<uuid>",
-  "type": "founder.opinion",
-  "labels": ["human_view"],
+  "type": "human_subjective_view.people",
   "subject": {"person": "Founder name"},
-  "value": {"speaker": "user", "view": "strong product taste"},
+  "value": {
+    "speaker": "<Llama user>",
+    "rawText": "I think the founder has strong product taste.",
+    "summary": "Strong product taste, per the user."
+  },
   "origin": {
     "kind": "user",
     "originalUserUtterance": "I think the founder has strong product taste."

package/CHANGELOG.md

@@ -6,6 +6,28 @@ this project adheres to [Semantic Versioning](https://semver.org).
 
 ## [Unreleased]
 
+## [2.2.0] - 2026-09-04
+
+### Added
+- Define `human_subjective_view.people` and `human_subjective_view.business`
+  as the only Information types for a Llama user's own judgment. CLI and MCP
+  reject them locally unless `value.speaker` is set and `value.rawText`
+  appears verbatim inside `origin.originalUserUtterance`; retired aliases
+  (`human_view`, `human_statement`, `human_opinion`, `human_judgment`,
+  `partner_view`) are refused with the replacement named. Core enforces the
+  same rule server-side.
+
+## [2.1.3] - 2026-09-03
+
+### Added
+- Add `llama page-schema list|read|section` and the matching
+  `get_live_deal_page_schema` MCP tool so Agents progressively load only the
+  exact Live Deal Page field contracts needed for a write.
+
+### Changed
+- Make Agent bootstrap and briefing request the compact Live Deal Page schema
+  index instead of downloading and repeating the complete field contract.
+
 ## [2.1.2] - 2026-09-03
 
 ### Fixed

package/README.md

@@ -180,7 +180,9 @@ Its Deal surface is also exactly four tools:
 Authentication, skill discovery, Wiki, admin audit, preferences, and external
 pitch remain separate non-Deal domains. Deal Memory likewise has two separate
 MCP tools, `get_deal_memory` and `update_deal_memory`; it does not add a fifth
-Occam Deal tool.
+Occam Deal tool. `get_live_deal_page_schema` progressively reads the Page
+schema index, exact fields, or one section; it is read-only context rather than
+a fifth Deal action.
 
 ## Agent bootstrap
 
@@ -192,10 +194,12 @@ llama skills show <slug>
 
 Agents should run `llama agent bootstrap` before Deal work. The command loads
 the authenticated, server-owned private Brain: Investment Framework V3 for
-thinking plus the Llama Command operating skill for working. It also loads the
-separate current Live Deal Page field contract for writing. None of that private
-content is shipped in this public package; CLI is only the tool and authenticated
-transport.
+thinking plus the Llama Command operating skill for working. It also loads a
+compact index of the current Live Deal Page fields. Before `page.patch`, use
+`llama page-schema read <field> [field...]` to load only the exact field
+contracts being changed, or `llama page-schema section <section>` when the
+write genuinely spans a section. None of that private content is shipped in
+this public package; CLI is only the tool and authenticated transport.
 
 The live server briefing is authoritative. The bundled
 `AGENT_BRIEFING.md` is an offline fallback with the same four-action contract.

package/README.zh-CN.md

@@ -170,12 +170,15 @@ MCP 的 Deal 工具同样恰好四个：
 认证、skill discovery、Wiki、admin audit、preferences 和 external pitch
 属于其他独立领域，不会扩大 Deal action space。Deal Memory 同样是独立领域，
 对应 `get_deal_memory` 和 `update_deal_memory` 两个 MCP 工具，并不增加第五个
-Occam Deal 工具。
+Occam Deal 工具。`get_live_deal_page_schema` 以渐进方式读取 Page schema 的
+索引、精确字段或一个 section；它是只读上下文，也不是第五个 Deal action。
 
 ## Agent 启动
 
 ```bash
 llama agent bootstrap
+llama page-schema list
+llama page-schema read <field> [field...]
 llama skills search "<任务>"
 llama skills show <slug>
 ```
@@ -183,8 +186,10 @@ llama skills show <slug>
 Agent 在处理 Deal 前应先运行 `llama agent bootstrap`。该命令会从鉴权后的
 Llama Command 服务端加载两部分私有 Brain：负责投资思考的 Investment
 Framework V3，以及负责如何在 Command 工作的 Llama Command operating skill；
-同时加载独立的、实时的 Live Deal Page 字段契约。所有私有内容都不会进入公开
-npm 包；CLI 只是 Agent 的工具和鉴权运输层。
+同时加载实时 Live Deal Page 字段的精简索引。执行 `page.patch` 前，通过
+`llama page-schema read <field> [field...]` 只加载本次修改所需的字段契约；
+只有当写入确实跨越一个 section 时才加载整个 section。所有私有内容都不会进入
+公开 npm 包；CLI 只是 Agent 的工具和鉴权运输层。
 
 服务端实时 briefing 是权威合同；包内 `AGENT_BRIEFING.md` 是相同四动作
 合同的离线兜底。

package/bin/llama-mcp.mjs

@@ -26,6 +26,7 @@ import {
   prepareDealCommand,
 } from "../lib/deal-actions.mjs";
 import { buildDealMemoryPath } from "../lib/deal-memory-actions.mjs";
+import { buildPageSchemaPath, MAX_PAGE_SCHEMA_FIELDS } from "../lib/page-schema.mjs";
 
 const requireFromHere = createRequire(import.meta.url);
 const { version: PKG_VERSION } = requireFromHere("../package.json");
@@ -151,7 +152,7 @@ server.registerTool(
 server.registerTool(
   "write_deal",
   {
-    description: "The only Deal mutation tool. Pass one command: input.submit, information.put, page.patch, or artifact.put. All Page content fields are Agent-writable; author controls attribution, not permission. Use the bootstrap field contract and bilingual Page prose. page.patch is JSON Merge Patch: arrays replace whole arrays, so read-modify-write and preserve sibling slots.",
+    description: "The only Deal mutation tool. Pass one command: input.submit, information.put, page.patch, or artifact.put. All Page content fields are Agent-writable; author controls attribution, not permission. Before page.patch, use get_live_deal_page_schema for the exact fields being changed and write bilingual Page prose. page.patch is JSON Merge Patch: arrays replace whole arrays, so read-modify-write and preserve sibling slots. A Llama user's own judgment is human_subjective_view.people or human_subjective_view.business and must carry value.speaker plus value.rawText quoted verbatim from origin.originalUserUtterance; it is rejected otherwise.",
     // MCP SDK 1.30 only publishes object schemas at the tool root. Nesting one
     // command preserves the real discriminated union instead of flattening four
     // incompatible operations into false affordances.
@@ -216,11 +217,12 @@ server.registerTool(
 server.registerTool(
   "agent_bootstrap",
   {
-    description: "Mandatory before Llama work: fetch the authenticated two-part private Brain (Investment Framework V3 + Llama Command operating skill), the separate current Live Deal Page field contract, runtime contract, and visible Llama OS skills.",
+    description: "Mandatory before Llama work: fetch the authenticated two-part private Brain (Investment Framework V3 + Llama Command operating skill), a compact Live Deal Page schema index, runtime contract, and visible Llama OS skills. Read exact Page fields separately with get_live_deal_page_schema.",
     inputSchema: { limit: z.number().int().min(1).max(100).optional() },
   },
   async ({ limit } = {}) => {
     const params = new URLSearchParams({ clientVersion: PKG_VERSION });
+    params.set("page_schema", "progressive");
     if (limit) params.set("limit", String(limit));
     try {
       const result = await request("GET", `/api/agent/manifest?${params}`);
@@ -231,6 +233,25 @@ server.registerTool(
   },
 );
 
+server.registerTool(
+  "get_live_deal_page_schema",
+  {
+    description: "Read the Live Deal Page schema progressively. Call with no selector for the compact index; before page.patch, request only exact fields or one section. This is read-only context, not a fifth Deal action.",
+    inputSchema: {
+      fields: z.array(z.string().min(1)).min(1).max(MAX_PAGE_SCHEMA_FIELDS).optional(),
+      section: z.string().min(1).optional(),
+    },
+  },
+  async ({ fields, section } = {}) => {
+    try {
+      // @core-api-operation GET /api/agent/page-schema
+      return await callApi("GET", buildPageSchemaPath({ fields, section }));
+    } catch (error) {
+      return textResult(formatErrorForDisplay(error), true);
+    }
+  },
+);
+
 server.registerTool(
   "skills_search",
   {
@@ -430,7 +451,7 @@ server.registerTool(
 server.registerPrompt(
   "agent_briefing",
   {
-    description: "Mandatory before Llama work: fetch the authenticated two-part private Brain, the separate current Live Deal Page field contract, and runtime contract; bundled text is only an offline fallback.",
+    description: "Mandatory before Llama work: fetch the authenticated two-part private Brain, compact Live Deal Page schema index, and runtime contract; read exact fields separately with get_live_deal_page_schema. Bundled text is only an offline fallback.",
   },
   async () => {
     const headers = await getAuthHeaders();
@@ -440,6 +461,7 @@ server.registerPrompt(
     } else {
       try {
         const params = new URLSearchParams({ clientVersion: PKG_VERSION });
+        params.set("page_schema", "progressive");
         const response = await request("GET", `/api/agent/briefing?${params}`);
         text = response?.briefing || readBriefing();
       } catch (error) {

package/bin/llama.mjs

@@ -43,6 +43,7 @@ import {
   buildDealMemoryPath,
   readMarkdownInput,
 } from "../lib/deal-memory-actions.mjs";
+import { buildPageSchemaPath } from "../lib/page-schema.mjs";
 
 const requireFromHere = createRequire(import.meta.url);
 const { version: PKG_VERSION } = requireFromHere("../package.json");
@@ -65,6 +66,7 @@ Separate preserved domains:
   llama auth status|login|logout
   llama token set|show
   llama agent bootstrap
+  llama page-schema list|read|section
   llama skills list|search|show
   llama pref list|add|approve|retire
   llama explain <url-or-object>
@@ -84,14 +86,16 @@ Exactly four actions:
 
 Before any Page write:
   llama agent bootstrap
+  llama page-schema read <field> [field...]
 
 The authenticated bootstrap contains the two-part private Llama Brain:
 Investment Framework V3 plus the Llama Command operating skill. It also
-contains the separate current Page field contract; its field count is live,
-so read it there. CLI is the tool, not Brain. Information never updates
-Page automatically. page.patch is JSON
-Merge Patch: objects merge, null deletes, and arrays replace whole arrays, so
-read-modify-write complete Page arrays and preserve sibling slots.
+contains a compact index of the current Page fields. Before page.patch, use
+the separate page-schema command to read only the exact fields being changed,
+or one section when necessary. CLI is the tool, not Brain. Information never
+updates Page automatically. page.patch is JSON Merge Patch: objects merge,
+null deletes, and arrays replace whole arrays, so read-modify-write complete
+Page arrays and preserve sibling slots.
 
 create JSON:
   {"companyName":"Acme","page":{},"information":[],"origin":{"kind":"user","originalUserUtterance":"..."}}
@@ -99,6 +103,7 @@ create JSON:
 write JSON (CLI derives idempotencyKey when omitted):
   input.submit:    {"operation":"input.submit","dealId":"<uuid>","content":"<original input>","format":"text","source":{},"origin":{"kind":"user","originalUserUtterance":"..."}}
   information.put:{"operation":"information.put","dealId":"<uuid>","type":"founder_claim","labels":["founder-stated"],"subject":{"topic":"traction"},"value":{"content":"...","source":{"kind":"first_meeting_transcript"}},"origin":{"kind":"user","originalUserUtterance":"..."}}
+  human judgment:  {"operation":"information.put","dealId":"<uuid>","type":"human_subjective_view.people","subject":{"topic":"founder_execution"},"value":{"speaker":"<Llama user>","rawText":"<the user's verbatim words>","summary":"<optional agent summary>"},"origin":{"kind":"user","originalUserUtterance":"<complete user input; must contain rawText>"}}
   page.patch:      {"operation":"page.patch","dealId":"<uuid>","patch":{"description":{"en":"...","zh":"..."}},"origin":{"kind":"agent"}}
   artifact.put:    {"operation":"artifact.put","dealId":"<uuid>","kind":"deck","title":"deck.pdf","mimeType":"application/pdf","contentBase64":"<file bytes, base64>","origin":{"kind":"user","originalUserUtterance":"..."}}
 
@@ -111,6 +116,12 @@ with storageKey (+ optional storageUrl) plus byteSize and sha256 instead.
 For Information, origin records who caused the write and preserves user words;
 value.source records where the evidence came from.
 
+human_subjective_view.people|business is a Llama user's own judgment about the
+team or the deal. Subjective = the truth depends on who judges: a value or
+comparative word with no benchmark, or a claim about someone's future behavior
+or inner state. It requires value.speaker and value.rawText quoted verbatim from
+origin.originalUserUtterance; CLI and Core reject anything else. Other types are free.
+
 Core owns Chat Records, append-only Deal Events, Drive provisioning, audit,
 and idempotency. User-originated work must preserve exact wording in
 origin.originalUserUtterance or reference origin.originatingChatRecordId.`;
@@ -142,6 +153,16 @@ read this Story.
 This uses your existing Llama authentication. The CLI talks only to Command
 Core and never receives Deal Memory service credentials or AWS access.`;
 
+const HELP_PAGE_SCHEMA = `Llama Live Deal Page schema — progressive, read-only context
+
+  llama page-schema list
+  llama page-schema read <field> [field...]
+  llama page-schema section <section>
+
+Start with list. Before page.patch, read only the exact fields being changed;
+use a section read only when the write genuinely spans that section. This is a
+schema/field-prompt surface, not a fifth Deal action and not Page data.`;
+
 function parseFlags(args, allowed = null) {
   const flags = {};
   const positional = [];
@@ -168,7 +189,12 @@ function parseFlags(args, allowed = null) {
 }
 
 function usage(area) {
-  console.log(area === "deal" ? HELP_DEAL : area === "memory" ? HELP_MEMORY : HELP_ROOT);
+  console.log(
+    area === "deal" ? HELP_DEAL :
+    area === "memory" ? HELP_MEMORY :
+    area === "page-schema" ? HELP_PAGE_SCHEMA :
+    HELP_ROOT,
+  );
 }
 
 function assertDealAction(area, action) {
@@ -502,6 +528,7 @@ async function main() {
     }
     try {
       const params = new URLSearchParams({ clientVersion: PKG_VERSION });
+      params.set("page_schema", "progressive");
       const response = await request("GET", `/api/agent/briefing?${params}`);
       process.stdout.write(response?.briefing || readBriefing());
     } catch (error) {
@@ -514,6 +541,7 @@ async function main() {
   if (area === "agent" && action === "bootstrap") {
     const { flags } = parseFlags(rest, ["json", "limit"]);
     const params = new URLSearchParams({ clientVersion: PKG_VERSION });
+    params.set("page_schema", "progressive");
     if (flags.limit && flags.limit !== true) params.set("limit", String(flags.limit));
     const manifest = await request("GET", `/api/agent/manifest?${params}`);
     if (flags.json) print(structuredBootstrapManifest(manifest));
@@ -521,6 +549,28 @@ async function main() {
     return;
   }
 
+  if (area === "page-schema") {
+    if (!action || action === "list") {
+      if (rest.length) throw new Error("Usage: llama page-schema list");
+      // @core-api-operation GET /api/agent/page-schema
+      print(await request("GET", buildPageSchemaPath()));
+      return;
+    }
+    if (action === "read") {
+      if (!rest.length) throw new Error("Usage: llama page-schema read <field> [field...]");
+      // @core-api-operation GET /api/agent/page-schema
+      print(await request("GET", buildPageSchemaPath({ fields: rest })));
+      return;
+    }
+    if (action === "section") {
+      if (rest.length !== 1) throw new Error("Usage: llama page-schema section <section>");
+      // @core-api-operation GET /api/agent/page-schema
+      print(await request("GET", buildPageSchemaPath({ section: rest[0] })));
+      return;
+    }
+    throw new Error("Page schema actions: list, read, section.");
+  }
+
   if (area === "skills" || (area === "agent" && action === "skills")) {
     const sub = area === "skills" ? action : rest[0];
     const args = area === "skills" ? rest : rest.slice(1);

package/contracts/core-api.json

@@ -1,9 +1,9 @@
 {
   "format": "llama.core-api-contract.v1",
   "name": "llama-core-api",
-  "apiVersion": "5.1.0",
+  "apiVersion": "5.3.0",
   "openapiVersion": "3.0.3",
-  "sha256": "948a8d02ae833e8cfa2302bc3c615fc1d3c792ae774ebc25b16be81fd1b85fee",
-  "pathCount": 111,
-  "operationCount": 138
+  "sha256": "b793f67038b48cdfbfcc9141153132689c40e84adfbe087b0dbef3a409f5e5d5",
+  "pathCount": 113,
+  "operationCount": 140
 }

package/contracts/required-operations.json

@@ -9,6 +9,7 @@
     { "method": "GET", "path": "/api/agent/briefing" },
     { "method": "GET", "path": "/api/agent/explain" },
     { "method": "GET", "path": "/api/agent/manifest" },
+    { "method": "GET", "path": "/api/agent/page-schema" },
     { "method": "GET", "path": "/api/agent/preferences" },
     { "method": "GET", "path": "/api/agent/skills" },
     { "method": "GET", "path": "/api/agent/skills/{slug}" },

package/lib/build-manifest.json

@@ -1,15 +1,15 @@
 {
   "format": "llama.cli-build.v1",
   "packageName": "@llamaventures/cli",
-  "packageVersion": "2.1.2",
-  "sourceSha": "bc642686f5c52df26bd927f7829303e5b10d3ce7",
+  "packageVersion": "2.2.0",
+  "sourceSha": "2f4c73310ed6c5d36a47a275490ff1082321ca27",
   "sourceKind": "github",
   "sourceDirty": false,
   "coreApiContract": {
     "format": "llama.core-api-contract.v1",
     "name": "llama-core-api",
-    "apiVersion": "5.1.0",
+    "apiVersion": "5.3.0",
     "openapiVersion": "3.0.3",
-    "sha256": "948a8d02ae833e8cfa2302bc3c615fc1d3c792ae774ebc25b16be81fd1b85fee"
+    "sha256": "b793f67038b48cdfbfcc9141153132689c40e84adfbe087b0dbef3a409f5e5d5"
   }
 }

package/lib/deal-actions.mjs

@@ -9,6 +9,72 @@ export const DEAL_WRITE_OPERATIONS = Object.freeze([
   "artifact.put",
 ]);
 
+/**
+ * A Llama user's own subjective judgment about the team (.people) or the deal
+ * (.business). An agent may record it, but the record must name the user and
+ * quote them verbatim from their own input, so the type alone stays a reliable
+ * filter. Core enforces the same rule; this check only fails faster.
+ */
+export const HUMAN_SUBJECTIVE_VIEW_TYPES = Object.freeze([
+  "human_subjective_view.people",
+  "human_subjective_view.business",
+]);
+
+const RETIRED_HUMAN_VIEW_TYPES = Object.freeze([
+  "human_view",
+  "human_statement",
+  "human_opinion",
+  "human_judgment",
+  "partner_view",
+]);
+
+function normalizeWhitespace(text) {
+  return String(text).replace(/\s+/g, " ").trim();
+}
+
+function stringField(object, key) {
+  const value = object && typeof object === "object" && !Array.isArray(object) ? object[key] : undefined;
+  return typeof value === "string" ? value.trim() : "";
+}
+
+export function validateHumanSubjectiveView(information, origin) {
+  const type = stringField(information, "type");
+  const lower = type.toLowerCase();
+  if (RETIRED_HUMAN_VIEW_TYPES.includes(lower) || lower.startsWith("human_view.")) {
+    throw new Error(
+      `Information type "${type}" is retired: a Llama user's own judgment is human_subjective_view.people or human_subjective_view.business`,
+    );
+  }
+  if (lower !== "human_subjective_view" && !lower.startsWith("human_subjective_view.")) {
+    return information;
+  }
+  if (!HUMAN_SUBJECTIVE_VIEW_TYPES.includes(type)) {
+    throw new Error(
+      "human_subjective_view must be exactly human_subjective_view.people (team) or human_subjective_view.business (deal)",
+    );
+  }
+  const utterance = origin?.kind === "user" ? stringField(origin, "originalUserUtterance") : "";
+  if (!utterance) {
+    throw new Error(
+      "human_subjective_view.* must originate from a user and carry the user's words in origin.originalUserUtterance",
+    );
+  }
+  const speaker = stringField(information.value, "speaker");
+  const rawText = stringField(information.value, "rawText");
+  if (!speaker) {
+    throw new Error("human_subjective_view.* requires value.speaker: the Llama user whose judgment this is");
+  }
+  if (!rawText) {
+    throw new Error("human_subjective_view.* requires value.rawText: the user's verbatim words");
+  }
+  if (!normalizeWhitespace(utterance).includes(normalizeWhitespace(rawText))) {
+    throw new Error(
+      "human_subjective_view.* value.rawText must appear verbatim inside origin.originalUserUtterance; do not paraphrase",
+    );
+  }
+  return information;
+}
+
 function canonicalJson(value) {
   if (Array.isArray(value)) return `[${value.map(canonicalJson).join(",")}]`;
   if (value && typeof value === "object") {
@@ -153,9 +219,14 @@ export function prepareDealCommand(action, input) {
   }
   if (action === "create") {
     validateLocalizedTextPairs(command.page ?? {});
+    const information = Array.isArray(command.information) ? command.information : [];
+    for (const item of information) validateHumanSubjectiveView(item, command.origin);
   }
   if (action === "write" && command.operation === "page.patch") {
     validateLocalizedTextPairs(command.patch ?? {});
   }
+  if (action === "write" && command.operation === "information.put") {
+    validateHumanSubjectiveView(command, command.origin);
+  }
   return ensureIdempotencyKey(validateCommandOrigin(command));
 }

package/lib/page-schema.mjs

@@ -0,0 +1,24 @@
+export const LIVE_DEAL_PAGE_SCHEMA_PATH = "/api/agent/page-schema";
+export const MAX_PAGE_SCHEMA_FIELDS = 20;
+
+export function buildPageSchemaPath({ fields = [], section } = {}) {
+  const normalizedFields = [...new Set(
+    fields.map((field) => String(field).trim()).filter(Boolean),
+  )];
+  const normalizedSection = section === undefined ? "" : String(section).trim();
+
+  if (normalizedFields.length && normalizedSection) {
+    throw new Error("Choose exact fields or one section, not both.");
+  }
+  if (normalizedFields.length > MAX_PAGE_SCHEMA_FIELDS) {
+    throw new Error(`Choose at most ${MAX_PAGE_SCHEMA_FIELDS} exact fields per request.`);
+  }
+  if (section !== undefined && !normalizedSection) {
+    throw new Error("Page schema section cannot be empty.");
+  }
+
+  const params = new URLSearchParams();
+  for (const field of normalizedFields) params.append("field", field);
+  if (normalizedSection) params.set("section", normalizedSection);
+  return `${LIVE_DEAL_PAGE_SCHEMA_PATH}${params.size ? `?${params}` : ""}`;
+}

package/package.json

@@ -1,12 +1,12 @@
 {
   "name": "@llamaventures/cli",
-  "version": "2.1.2",
+  "version": "2.2.0",
   "description": "CLI + MCP server for the Llama Ventures investment workbench (command.llamaventures.vc).",
   "type": "module",
   "scripts": {
     "test": "npm run test:agent-routing && npm run test:contract",
     "test:agent-routing": "node scripts/verify-agent-routing.mjs",
-    "test:contract": "node --test scripts/build-manifest.test.mjs scripts/core-api-call-sites.test.mjs scripts/deal-actions-cli.test.mjs scripts/deal-actions.test.mjs scripts/deal-memory-actions-cli.test.mjs scripts/deal-memory-actions.test.mjs scripts/mcp-four-action-surface.test.mjs scripts/server-compatibility.test.mjs scripts/verify-core-api-contract.test.mjs && node scripts/verify-core-api-contract.mjs",
+    "test:contract": "node --test scripts/build-manifest.test.mjs scripts/core-api-call-sites.test.mjs scripts/deal-actions-cli.test.mjs scripts/deal-actions.test.mjs scripts/deal-memory-actions-cli.test.mjs scripts/deal-memory-actions.test.mjs scripts/mcp-four-action-surface.test.mjs scripts/page-schema-cli.test.mjs scripts/page-schema.test.mjs scripts/server-compatibility.test.mjs scripts/verify-core-api-contract.test.mjs && node scripts/verify-core-api-contract.mjs",
     "verify:artifact": "node scripts/verify-release-artifact.mjs",
     "verify:release": "npm test && npm run verify:artifact && node scripts/verify-tarball-clean.mjs",
     "prepack": "node scripts/prepare-build-manifest.mjs",