agentpage 0.0.50 → 0.0.52

package/dist/index.mjs

@@ -550,6 +550,8 @@ function buildCompactMessages(userMessage, trace, latestSnapshot, currentUrl, hi
 	});
 	const hasErrors = trace.some((e) => hasToolError(e.result));
 	const contextParts = [
+		`Original Goal: ${userMessage}`,
+		"",
 		`Remaining: ${activeInstruction}`,
 		"",
 		"Batch fills per round; clicks end the round — at most ONE click (last). Do NOT call page_info (snapshot is auto-refreshed).",
@@ -736,6 +738,45 @@ function detectIdleLoop(toolCalls, consecutiveReadOnlyRounds) {
 	}
 	return 0;
 }
+/**
+* 重复无效点击拦截。
+*
+* 场景：模型反复点击同一个 selector 但页面快照从未变化。
+* 框架在每轮结束时通过快照指纹对比发现点击无效后，将 selector 加入
+* `ineffectiveClickSelectors` 集合。下一轮模型再次点击相同 selector 时，
+* 直接拦截并返回错误提示，引导模型换目标。
+*
+* 使用方式：
+* - 在工具执行前调用此函数
+* - 仅对 `dom.click` 动作生效
+* - 返回 null 表示放行，返回 ToolCallResult 表示拦截
+*
+* 集合的维护由 index.ts 负责：
+* - 轮次结束且快照未变：将本轮 click 的 selector 加入集合
+* - 快照变化：清空集合（页面已改变，之前的判定失效）
+*/
+function checkIneffectiveClickRepeat(toolName, toolInput, ineffectiveClickSelectors) {
+	if (toolName !== "dom") return null;
+	if (getToolAction(toolInput) !== "click") return null;
+	const selector = toolInput && typeof toolInput === "object" ? toolInput.selector : void 0;
+	if (typeof selector !== "string" || !selector) return null;
+	if (!ineffectiveClickSelectors.has(selector)) return null;
+	return {
+		content: [
+			`Click on ${selector} was BLOCKED — this target was already clicked in a previous round with NO visible effect on the page.`,
+			"You MUST try a DIFFERENT element. Suggestions:",
+			"1) Look INSIDE the clicked container for an <a>, <button>, or child with clk/pdn/mdn listener",
+			"2) Try a parent or sibling element with stronger click signal",
+			"3) Use evaluate to inspect or trigger navigation programmatically",
+			"4) Try a completely different approach (search, sidebar, direct URL navigation)"
+		].join("\n"),
+		details: {
+			error: true,
+			code: "INEFFECTIVE_CLICK_BLOCKED",
+			selector
+		}
+	};
+}
 
 //#endregion
 //#region src/core/agent-loop/index.ts
@@ -839,6 +880,7 @@ async function executeAgentLoop(params) {
 	};
 	let recoveryCount = 0;
 	let redundantInterceptCount = 0;
+	const ineffectiveClickSelectors = /* @__PURE__ */ new Set();
 	let pendingNotFoundRetry;
 	let snapshotReadCount = 0;
 	let snapshotSizeTotal = 0;
@@ -1023,6 +1065,7 @@ async function executeAgentLoop(params) {
 		let roundHasConfirmedProgress = false;
 		const executedTaskCalls = [];
 		const roundMissingTasks = [];
+		const roundClickSelectors = [];
 		for (const tc of response.toolCalls) {
 			if (tc.name === "dom" && getToolAction(tc.input) === "scroll") {
 				const ref = extractHashSelectorRef(tc.input);
@@ -1035,6 +1078,15 @@ async function executeAgentLoop(params) {
 				callbacks?.onToolResult?.(tc.name, redundant);
 				continue;
 			}
+			const ineffective = checkIneffectiveClickRepeat(tc.name, tc.input, ineffectiveClickSelectors);
+			if (ineffective) {
+				appendToolTrace(round, tc.name, tc.input, ineffective);
+				redundantInterceptCount += 1;
+				callbacks?.onToolCall?.(tc.name, tc.input);
+				callbacks?.onToolResult?.(tc.name, ineffective);
+				roundHasError = true;
+				continue;
+			}
 			callbacks?.onToolCall?.(tc.name, tc.input);
 			let result = await registry.dispatch(tc.name, tc.input);
 			const debounced = applySnapshotDebounce(tc.name, tc.input, result, consecutiveSnapshotCalls);
@@ -1048,6 +1100,10 @@ async function executeAgentLoop(params) {
 				name: tc.name,
 				input: tc.input
 			});
+			if (tc.name === "dom" && getToolAction(tc.input) === "click" && !hasToolError(result)) {
+				const sel = tc.input && typeof tc.input === "object" ? tc.input.selector : void 0;
+				if (typeof sel === "string" && sel) roundClickSelectors.push(sel);
+			}
 			const missingTask = collectMissingTask(tc.name, tc.input, result);
 			if (missingTask) roundMissingTasks.push(missingTask);
 			if (result.details && typeof result.details === "object") roundHasError = roundHasError || Boolean(result.details.error);
@@ -1106,7 +1162,11 @@ async function executeAgentLoop(params) {
 					"- Look INSIDE the target for <a>/<button>/child with clk listener, or try a parent/sibling with stronger signal, or use a completely different approach."
 				].join("\n");
 				protocolViolationHint = protocolViolationHint ? protocolViolationHint + "\n\n" + unchangedHint : unchangedHint;
-			} else if (roundEndFingerprint !== roundStartFingerprint) consecutiveNoProtocolRounds = 0;
+				for (const sel of roundClickSelectors) ineffectiveClickSelectors.add(sel);
+			} else if (roundEndFingerprint !== roundStartFingerprint) {
+				consecutiveNoProtocolRounds = 0;
+				ineffectiveClickSelectors.clear();
+			}
 		}
 		if (consecutiveNoProtocolRounds >= 5) {
 			finalReply = response.text?.trim() || "任务已完成。";
@@ -2273,6 +2333,8 @@ function buildSystemPrompt(params = {}) {
 		"You are AutoPilot, an AI agent controlling the current web page via tools.",
 		"",
 		"## Core Rules",
+		"- **Original Goal Anchor:** The user's original input is provided as `Original Goal` every round. Your plan and each action must NEVER deviate from it. If the current page shows a 'Create X' button but the user said 'go to X', you must navigate INTO existing X, NOT create a new one.",
+		"- **Goal decomposition:** Distinguish the TARGET entity from the ACTION to perform. 'go to X and do Y' = locate X → enter X → do Y inside X. 'create X' = make a new X. 'edit X' = find existing X → modify it. Never confuse navigating to an entity with creating/deleting/modifying it. If the target entity is not visible, search or filter for it first — do not pick the nearest similarly-named button.",
 		"- Work from CURRENT snapshot + remaining task. Do not restate.",
 		"- Task reduction: (remaining, prev actions, this-round) → new remaining.",
 		"- Use #hashID from snapshot as selector. Do not guess CSS selectors.",