@webskill/sdk 0.6.0 → 0.7.0

package/dist/browser.js

@@ -1,5 +1,5 @@
 import { A as parseSkillMarkdown, B as unzipWithLimits, H as verifyManifest, L as resolveInsideRoot, M as readResponseWithLimit, N as readSkillSignature, O as messageOf, T as isValidSkillName, U as verifySkillSignature, V as validateSkills, _ as atomicWriteText, f as SkillDiscovery, h as assertRemoteUrlAllowed, j as parseSkillPackManifest, m as WebSkillError, o as SKILLS_LOCKFILE, r as MANIFEST_EXCLUDED_FILES, s as SKILL_MANIFEST_FILE, u as SKILL_PACK_FILE, v as buildCatalog, w as exportSkills, y as buildManifest } from "./dist-8oQRa8Xz.js";
-import { H as createWebSkillApi, S as ProgressiveRouter, a as AgentLoop, at as normalizeToolContent, c as CapabilityApproval, et as mergeCatalogEntries, h as FsRunSnapshotStore, m as FsMemoryStore, nt as networkPolicyLibSource, o as AnthropicClient, ot as normalizeToolError, p as FsArtifactStore, st as parseBridgeRequest, x as OpenAiCompatibleClient, y as GoogleGenAiClient, z as bridgeError } from "./dist-DusANsrn.js";
+import { B as bridgeError, C as ProgressiveRouter, S as OpenAiCompatibleClient, U as createWebSkillApi, a as AgentLoop, b as GoogleGenAiClient, c as CapabilityApproval, ct as normalizeToolError, g as FsRunSnapshotStore, h as FsMemoryStore, it as networkPolicyLibSource, lt as parseBridgeRequest, m as FsArtifactStore, nt as mergeCatalogEntries, o as AnthropicClient, st as normalizeToolContent } from "./dist-D0qW6e40.js";
 import { n as MockLlmClient } from "./testing-CYTFqkDm.js";
 
 //#region ../browser/dist/index.js
@@ -1615,8 +1615,20 @@ function describeError(e) {
 	if (e instanceof TypeError) return "the request was blocked, most likely by CORS";
 	return e instanceof Error ? e.message : String(e);
 }
-/** 值可能是凭据的输入类型：即便宿主忘了写进 exclude 也不读值 */
+/** 值可能是凭据的输入类型：即便宿主忘了写进 exclude 也不读值。操作侧复用同一份，不另写 */
 const SECRET_INPUT_TYPES = /* @__PURE__ */ new Set(["password", "hidden"]);
+/** 可被操作的角色：只给这些节点发句柄，其余节点模型连提都提不出来 */
+const ACTIONABLE_ROLES = /* @__PURE__ */ new Set([
+	"button",
+	"checkbox",
+	"combobox",
+	"form",
+	"link",
+	"radio",
+	"searchbox",
+	"switch",
+	"textbox"
+]);
 /** tag → 可访问性角色的最小映射；命中不了就退到 generic，不猜 */
 const ROLE_BY_TAG = {
 	A: "link",
@@ -1654,7 +1666,7 @@ const clip = (text) => {
 	const normalized = text.replace(/\s+/g, " ").trim();
 	return normalized.length > MAX_TEXT ? `${normalized.slice(0, MAX_TEXT)}…` : normalized;
 };
-function roleOf(element) {
+function roleOf$1(element) {
 	const explicit = element.getAttribute("role");
 	if (explicit !== null && explicit.trim() !== "") return explicit.trim();
 	if (element.tagName === "INPUT") {
@@ -1700,7 +1712,7 @@ function valueOf(element) {
 * 隐藏节点不进结果：屏幕上看不见的东西，用户没有机会发现它被读走了。
 * jsdom 里 `getComputedStyle` 可用但布局是假的，因此只看显式声明，不看尺寸。
 */
-function hidden(element, view) {
+function hidden$1(element, view) {
 	if (element.hasAttribute("hidden")) return true;
 	if (element.getAttribute("aria-hidden") === "true") return true;
 	const inlineStyle = element.getAttribute("style") ?? "";
@@ -1718,18 +1730,18 @@ const SKIPPED_TAGS = /* @__PURE__ */ new Set([
 	"NOSCRIPT",
 	"TEMPLATE"
 ]);
-function describe(element, excluded, doc, view, targets) {
+function describe(element, excluded, doc, view, targets, handles) {
 	if (excluded.has(element)) return void 0;
 	if (SKIPPED_TAGS.has(element.tagName)) return void 0;
-	if (hidden(element, view)) return void 0;
+	if (hidden$1(element, view)) return void 0;
 	const children = [];
 	for (const child of element.children) {
-		const node = describe(child, excluded, doc, view, targets);
+		const node = describe(child, excluded, doc, view, targets, handles);
 		if (node !== void 0) children.push(node);
 	}
 	const name = nameOf(element, doc);
 	const value = valueOf(element);
-	const role = roleOf(element);
+	const role = roleOf$1(element);
 	if (role === "generic" && name === void 0 && value === void 0 && children.length === 0) return void 0;
 	const node = {
 		role,
@@ -1737,6 +1749,8 @@ function describe(element, excluded, doc, view, targets) {
 		...value !== void 0 ? { value } : {},
 		...children.length > 0 ? { children } : {}
 	};
+	const ref = handles?.issue(element, role);
+	if (ref !== void 0) node.ref = ref;
 	if (targets !== void 0 && isCapturableElement(element)) targets.push({
 		element,
 		node
@@ -1744,6 +1758,31 @@ function describe(element, excluded, doc, view, targets) {
 	return node;
 }
 /**
+* 句柄发放与解析。表在 reader 实例内、每次 `read()` 整体替换：
+* 全局表是第二套状态且会跨会话泄漏；不替换则页面变了还能用旧句柄点到别的东西。
+*/
+var HandleIssuer = class {
+	#table = /* @__PURE__ */ new Map();
+	#actionable;
+	#excluded;
+	constructor(actionable, excluded) {
+		this.#actionable = actionable;
+		this.#excluded = excluded;
+	}
+	issue(element, role) {
+		if (!this.#actionable.has(element) || this.#excluded.has(element)) return void 0;
+		if (!ACTIONABLE_ROLES.has(role)) return void 0;
+		const bytes = /* @__PURE__ */ new Uint8Array(8);
+		crypto.getRandomValues(bytes);
+		const ref = [...bytes].map((b) => b.toString(16).padStart(2, "0")).join("");
+		this.#table.set(ref, element);
+		return ref;
+	}
+	get table() {
+		return this.#table;
+	}
+};
+/**
 * DOM 遍历实现（设计 09 §2 的「实现层」）。产出只有角色 / 名称 / 值，
 * **不含任何 HTML 标记、class、data 属性或注释**（FR-10.4）。
 *
@@ -1752,17 +1791,43 @@ function describe(element, excluded, doc, view, targets) {
 * @experimental
 */
 function createDomPerceptionReader(options = {}) {
+	let handles;
+	const resolveDocument = () => options.document ?? globalThis.document;
+	/** 按当前 DOM 重算一次可操作集合；发句柄与执行前各算一次 */
+	const actionSets = (doc) => {
+		const actionable = /* @__PURE__ */ new Set();
+		const excluded = /* @__PURE__ */ new Set();
+		const scope = options.actionScope;
+		if (scope === void 0 || scope.include.length === 0) return {
+			actionable,
+			excluded
+		};
+		for (const selector of scope.include) for (const root of doc.querySelectorAll(selector)) {
+			actionable.add(root);
+			for (const descendant of root.querySelectorAll("*")) actionable.add(descendant);
+		}
+		for (const selector of scope.exclude ?? []) for (const root of doc.querySelectorAll(selector)) {
+			excluded.add(root);
+			for (const descendant of root.querySelectorAll("*")) excluded.add(descendant);
+		}
+		return {
+			actionable,
+			excluded
+		};
+	};
 	function collect(scope, targets) {
-		const doc = options.document ?? globalThis.document;
+		const doc = resolveDocument();
 		if (doc === void 0 || scope.include.length === 0) return [];
 		const view = doc.defaultView;
+		const sets = actionSets(doc);
+		handles = new HandleIssuer(sets.actionable, sets.excluded);
 		const excluded = /* @__PURE__ */ new Set();
 		for (const selector of scope.exclude ?? []) for (const element of doc.querySelectorAll(selector)) excluded.add(element);
 		const roots = [];
 		for (const selector of scope.include) for (const element of doc.querySelectorAll(selector)) if (!roots.includes(element)) roots.push(element);
 		const nodes = [];
 		for (const root of roots) {
-			const node = describe(root, excluded, doc, view, targets);
+			const node = describe(root, excluded, doc, view, targets, handles);
 			if (node !== void 0) nodes.push(node);
 		}
 		return nodes;
@@ -1804,7 +1869,126 @@ function createDomPerceptionReader(options = {}) {
 		if (!capture.images || capture.maxImages <= 0) return Promise.resolve({ nodes: collect(scope, void 0) });
 		return readWithImages(scope, capture);
 	}
-	return { read };
+	return {
+		read,
+		resolve: (ref) => handles?.table.get(ref),
+		inActionScope: (element) => {
+			const doc = resolveDocument();
+			if (doc === void 0) return false;
+			const sets = actionSets(doc);
+			return sets.actionable.has(element) && !sets.excluded.has(element);
+		}
+	};
+}
+const FILLABLE_TAGS = /* @__PURE__ */ new Set(["INPUT", "TEXTAREA"]);
+/** 只看显式声明，不看尺寸：jsdom 有 getComputedStyle 但没有布局 */
+function hidden(element) {
+	if (element.hasAttribute("hidden")) return true;
+	if (element.getAttribute("aria-hidden") === "true") return true;
+	const inlineStyle = element.getAttribute("style") ?? "";
+	if (/display\s*:\s*none|visibility\s*:\s*hidden/i.test(inlineStyle)) return true;
+	const view = element.ownerDocument.defaultView;
+	if (view !== null) {
+		const computed = view.getComputedStyle(element);
+		if (computed.display === "none" || computed.visibility === "hidden") return true;
+	}
+	return false;
+}
+const inputType = (element) => (element.getAttribute("type") ?? "text").toLowerCase();
+const isSecret = (element) => element.tagName === "INPUT" && SECRET_INPUT_TYPES.has(inputType(element));
+function accessibleName(element) {
+	const label = element.getAttribute("aria-label");
+	if (label !== null && label.trim() !== "") return label.trim();
+	const text = element.textContent?.replace(/\s+/g, " ").trim() ?? "";
+	if (text !== "") return text;
+	const placeholder = element.getAttribute("placeholder");
+	return placeholder !== null && placeholder.trim() !== "" ? placeholder.trim() : void 0;
+}
+function roleOf(element) {
+	const explicit = element.getAttribute("role");
+	if (explicit !== null && explicit.trim() !== "") return explicit.trim();
+	if (element.tagName === "INPUT") {
+		const type = inputType(element);
+		return type === "checkbox" || type === "radio" ? type : type === "button" || type === "submit" ? "button" : "textbox";
+	}
+	if (element.tagName === "BUTTON") return "button";
+	if (element.tagName === "A") return "link";
+	if (element.tagName === "TEXTAREA") return "textbox";
+	if (element.tagName === "SELECT") return "combobox";
+	if (element.tagName === "FORM") return "form";
+	return "generic";
+}
+const describeElement = (element) => {
+	const name = accessibleName(element);
+	return {
+		role: roleOf(element),
+		...name !== void 0 ? { name } : {},
+		...isSecret(element) ? { secret: true } : {}
+	};
+};
+/**
+* 受控组件必须收到 `input` / `change`：直接写 `.value` 在 React 上会被下一次渲染覆盖，
+* 表现为「填了又没了」。原生 setter 是绕过 React value 劫持的唯一办法。
+*/
+function fillValue(element, value) {
+	const prototype = element.tagName === "INPUT" ? HTMLInputElement.prototype : HTMLTextAreaElement.prototype;
+	const setter = Object.getOwnPropertyDescriptor(prototype, "value")?.set;
+	if (setter) setter.call(element, value);
+	else element.value = value;
+	element.dispatchEvent(new Event("input", { bubbles: true }));
+	element.dispatchEvent(new Event("change", { bubbles: true }));
+}
+/** @experimental */
+function createDomPageActionExecutor(options) {
+	const { reader } = options;
+	const describe = (ref) => {
+		const element = reader.resolve(ref);
+		return element === void 0 ? void 0 : describeElement(element);
+	};
+	const execute = (request) => {
+		const element = reader.resolve(request.ref);
+		if (element === void 0) return {
+			ok: false,
+			target: { role: "generic" },
+			reason: "The element reference is unknown or expired."
+		};
+		const target = describeElement(element);
+		const fail = (reason) => ({
+			ok: false,
+			target,
+			reason
+		});
+		if (!reader.inActionScope(element)) return fail("The element is no longer inside the actionable scope.");
+		if (hidden(element)) return fail("The element is not visible.");
+		if (element.hasAttribute("disabled")) return fail("The element is disabled.");
+		if (request.action === "click") {
+			element.click();
+			return {
+				ok: true,
+				target
+			};
+		}
+		if (request.action === "fill") {
+			if (!FILLABLE_TAGS.has(element.tagName)) return fail("The element is not a text control.");
+			if (element.hasAttribute("readonly")) return fail("The element is read-only.");
+			fillValue(element, request.value ?? "");
+			return {
+				ok: true,
+				target
+			};
+		}
+		const form = element.tagName === "FORM" ? element : element.form;
+		if (!form) return fail("The element does not belong to a form.");
+		form.requestSubmit();
+		return {
+			ok: true,
+			target
+		};
+	};
+	return {
+		execute,
+		describe
+	};
 }
 const isRecord = (v) => typeof v === "object" && v !== null;
 const isNonEmptyString = (v) => typeof v === "string" && v !== "";
@@ -2488,4 +2672,4 @@ async function deleteMemoryEncryptionKey(name = "user-profile") {
 }
 
 //#endregion
-export { BrowserSkillManager, BrowserWorkerScriptExecutor, ChromeBuiltinLlmClient, IframeWorkerLike, OpfsProvider, TsTranspiler, WORKER_BOOTSTRAP_SOURCE, WorkerRuntimeClient, WorkerUiBridge, bridgeError, captureElementImage, checkDictationAvailability, compressImageToBudget, createDomPerceptionReader, createEncryptedMemoryStore, createIframeWorker, createLlmClient, deleteMemoryEncryptionKey, extractZipWeb, generateMemoryEncryptionKey, installWebSkillNavigator, isCapturableElement, isCaptureFailure, isEncryptedMemoryValue, isOpfsAvailable, openMemoryEncryptionKey, parseBridgeRequest, parseMainMessage, parseWorkerEvent, probeChromeBuiltinAvailability, sha256HexWeb, startDictation, startWorkerRuntimeHost };
+export { BrowserSkillManager, BrowserWorkerScriptExecutor, ChromeBuiltinLlmClient, IframeWorkerLike, OpfsProvider, TsTranspiler, WORKER_BOOTSTRAP_SOURCE, WorkerRuntimeClient, WorkerUiBridge, bridgeError, captureElementImage, checkDictationAvailability, compressImageToBudget, createDomPageActionExecutor, createDomPerceptionReader, createEncryptedMemoryStore, createIframeWorker, createLlmClient, deleteMemoryEncryptionKey, extractZipWeb, generateMemoryEncryptionKey, installWebSkillNavigator, isCapturableElement, isCaptureFailure, isEncryptedMemoryValue, isOpfsAvailable, openMemoryEncryptionKey, parseBridgeRequest, parseMainMessage, parseWorkerEvent, probeChromeBuiltinAvailability, sha256HexWeb, startDictation, startWorkerRuntimeHost };