explorbot 0.1.32-beta.1 → 0.2.1

package/dist/src/ai/pilot.js

@@ -3,7 +3,6 @@ import dedent from 'dedent';
 import { z } from 'zod';
 import { ActionResult } from "../action-result.js";
 import { ConfigParser } from "../config.js";
-import { renderExperienceToc } from "../experience-tracker.js";
 import { TestResult } from "../test-plan.js";
 import { collectInteractiveNodes, detectFocusArea, extractFocusedElement } from "../utils/aria.js";
 import { ErrorPageError } from "../utils/error-page.js";
@@ -21,14 +20,18 @@ export class Pilot {
     conversation = null;
     researcher;
     explorer;
+    stateManager;
+    requestStore;
+    playwrightRecorder;
     fisherman = null;
-    experienceTracker;
-    constructor(provider, agentTools, researcher, explorer, experienceTracker) {
-        this.provider = provider;
+    constructor(deps, agentTools, researcher) {
+        this.provider = deps.ai;
         this.agentTools = agentTools;
         this.researcher = researcher;
-        this.explorer = explorer;
-        this.experienceTracker = experienceTracker || null;
+        this.explorer = deps.explorer;
+        this.stateManager = deps.stateManager;
+        this.requestStore = deps.requestStore;
+        this.playwrightRecorder = deps.playwrightRecorder;
     }
     setFisherman(fisherman) {
         this.fisherman = fisherman;
@@ -74,7 +77,7 @@ export class Pilot {
         let screenshotState = null;
         if (type === 'finish' && this.provider.hasVision()) {
             try {
-                screenshotState = await this.explorer.capturePageWithScreenshot();
+                screenshotState = await this.explorer.capture({ screenshot: true });
                 if (screenshotState.screenshot) {
                     visualAnalysis = (await this.researcher.answerQuestionAboutScreenshot(screenshotState, `Describe current page state relevant to: ${task.scenario}`)) || '';
                 }
@@ -90,7 +93,7 @@ export class Pilot {
             requestVerification: z
                 .string()
                 .nullable()
-                .describe('REQUIRED whenever decision is "pass" — a one-sentence natural-language claim about the current page that, if true, proves the scenario goal (e.g., "New test suite \\"Foo\\" is visible in the suites list"). NOT code: do not write I.*, expect(), .then(), grabTitle, or any JavaScript. Navigator translates the claim into CodeceptJS assertions and runs them; passing assertions are saved to the generated test file. Also use when evidence is insufficient before deciding pass/fail. Leave null for "continue", "fail", or "skipped".'),
+                .describe('REQUIRED whenever decision is "pass" — a one-sentence natural-language claim about the current page that, if true, proves the scenario goal (e.g., "New item \\"Foo\\" is visible in the items list"). NOT code: do not write I.*, expect(), .then(), grabTitle, or any JavaScript. Navigator translates the claim into CodeceptJS assertions and runs them; passing assertions are saved to the generated test file. Also use when evidence is insufficient before deciding pass/fail. Leave null for "continue", "fail", or "skipped".'),
         });
         const userContent = dedent `
       Tester wants to ${type} the test.
@@ -124,7 +127,7 @@ export class Pilot {
       - Mixed evidence + final state shows success → pass. Mixed + final state unclear → continue with guidance.
 
       When deciding "pass", you MUST also set requestVerification to a one-sentence natural-language
-      claim about the current page (e.g., "New test suite Foo is visible in the suites list"). NOT
+      claim about the current page (e.g., "New item Foo is visible in the items list"). NOT
       code — do not write I.*, expect(), .then(), or any JavaScript. Choose the strongest single
       piece of evidence (a unique element/text that exists ONLY because the scenario succeeded).
       Navigator translates the claim into CodeceptJS assertions; without it the generated test has
@@ -133,7 +136,7 @@ export class Pilot {
         const messages = [
             {
                 role: 'system',
-                content: this.buildVerdictSystemPrompt(type, task),
+                content: this.buildVerdictSystemPrompt(task),
             },
             { role: 'user', content: userContent },
         ];
@@ -151,7 +154,7 @@ export class Pilot {
                 tag('substep').log(`Pilot requesting verification: ${result.requestVerification}`);
                 const verifyResult = await navigator.verifyState(result.requestVerification, currentState).catch(() => null);
                 if (verifyResult?.verified && verifyResult.assertionSteps?.length) {
-                    this.explorer.getPlaywrightRecorder().recordVerification(verifyResult.assertionSteps);
+                    this.playwrightRecorder.recordVerification(verifyResult.assertionSteps);
                 }
             }
             tag('info').log(`Pilot: ${result.decision} - ${result.reason}`);
@@ -261,10 +264,8 @@ export class Pilot {
             return true;
         }
     }
-    buildSharedEvidenceRules(task) {
+    buildSharedEvidenceRules() {
         return dedent `
-      SCENARIO: ${task.scenario}
-
       EVIDENCE PRIORITY (strict):
       1) Final observable state proving the scenario goal
       2) verify()/see() results in the LAST few actions before stop/finish
@@ -297,6 +298,11 @@ export class Pilot {
 
       Expected results are MILESTONES, not the goal. Never fail because a milestone (toast, icon, styling)
       didn't match if the scenario goal IS accomplished.
+    `;
+    }
+    buildTaskContext(task) {
+        return dedent `
+      SCENARIO: ${task.scenario}
 
       ${this.buildDeletionScope(task)}
 
@@ -310,7 +316,7 @@ export class Pilot {
       iteration's work, but server-side side effects (records created, forms submitted) persist.
       Unnecessary resets create duplicate data and infinite loops.
 
-      ${this.buildSharedEvidenceRules(task)}
+      ${this.buildSharedEvidenceRules()}
 
       DECISION:
       - "allow": current page cannot host the scenario, irrecoverable error, or no path back.
@@ -327,16 +333,18 @@ export class Pilot {
       to verify, how to record. Do not suggest repeating actions that already succeeded.
       If progress is blocked only because the page lacks target data for the scenario, prefer precondition()
       over repeated UI attempts.
+
+      ${this.buildTaskContext(task)}
     `;
     }
-    buildVerdictSystemPrompt(type, task) {
+    buildVerdictSystemPrompt(task) {
         return dedent `
-      You are Pilot — final decision maker for test pass/fail. Tester requested ${type}. Review the
-      evidence and commit to a verdict; "continue" only when evidence is genuinely insufficient.
+      You are Pilot — final decision maker for test pass/fail. Review the evidence and commit to a
+      verdict; "continue" only when evidence is genuinely insufficient.
 
       ${capabilityGroundingRule}
 
-      ${this.buildSharedEvidenceRules(task)}
+      ${this.buildSharedEvidenceRules()}
 
       DECISION:
       - "pass": scenario goal is fully accomplished. Set requestVerification to a one-sentence claim about
@@ -355,6 +363,8 @@ export class Pilot {
       reason field: one short sentence, maximum 120 characters. Do NOT restate the decision
       ("scenario goal achieved/not achieved"). State what happened: what was verified, what failed,
       or what evidence was found.
+
+      ${this.buildTaskContext(task)}
     `;
     }
     async planTest(task, currentState) {
@@ -379,10 +389,6 @@ export class Pilot {
 
         FIRST: Decide if precondition() is needed.
 
-        ${capabilityGroundingRule}
-
-        ${dataProtectionRules}
-
         Call precondition() WHEN:
         - The scenario edits/deletes/modifies an item, and you want a DISPOSABLE item to act on safely
         - The scenario needs specific data clearly NOT on the current page (e.g., items with specific statuses for filtering)
@@ -425,8 +431,6 @@ export class Pilot {
             .filter((t) => t.wasSuccessful)
             .slice(-this.stepsToReview);
         const actionsContext = this.formatActions(toolCalls);
-        this.conversation.cleanupTag('page_summary', '...trimmed...', 1);
-        this.conversation.cleanupTag('recent_actions', '...trimmed...', 2);
         return this.sendToPilot(dedent `
         Navigated to new page.
         START URL: ${task.startUrl}
@@ -446,7 +450,7 @@ export class Pilot {
         ${this.formatExpectations(task)}
 
         First: evaluate whether this navigation makes sense for the scenario goal. If the page is unrelated, instruct Tester to back() or reset(). Then plan next steps.
-      `, 'pilot.reviewNewPage');
+      `, 'pilot.reviewNewPage', { task });
     }
     async analyzeProgress(task, currentState, testerConversation) {
         tag('substep').log('Pilot analyzing progress...');
@@ -458,7 +462,6 @@ export class Pilot {
         const toolCalls = testerConversation.getToolExecutions().slice(-this.stepsToReview);
         const actionsContext = this.formatActions(toolCalls);
         const stateContext = this.buildStateContext(currentState);
-        this.conversation.cleanupTag('recent_actions', '...trimmed...', 2);
         const hasFailures = toolCalls.length === 0 || toolCalls.some((t) => !t.wasSuccessful);
         const text = await this.sendToPilot(dedent `
         START URL: ${task.startUrl}
@@ -486,7 +489,7 @@ export class Pilot {
         const remaining = task.getRemainingExpectations();
         return `CHECKED: ${checked.length > 0 ? checked.join(', ') : 'none'}\nREMAINING: ${remaining.length > 0 ? remaining.join(', ') : 'none'}`;
     }
-    async sendToPilot(userText, functionId, opts = {}) {
+    async sendToPilot(userText, functionId, opts) {
         debugLog(`sendToPilot: ${functionId}, tools: ${!!opts.tools}, roundtrips: ${opts.maxToolRoundtrips ?? 0}`);
         let finalUserText = userText;
         if (opts.tools) {
@@ -496,17 +499,12 @@ export class Pilot {
             }
         }
         this.conversation.addUserText(finalUserText);
-        let tools;
-        if (opts.tools) {
-            tools = this.pickPlanningTools();
-        }
-        if (opts.tools && opts.task) {
-            tools = { ...tools, ...this.buildPreconditionTool(opts.task) };
-        }
+        const tools = { ...this.pickPlanningTools(), ...this.buildPreconditionTool(opts.task) };
         const result = await this.provider.invokeConversation(this.conversation, tools, {
             maxToolRoundtrips: opts.maxToolRoundtrips ?? 0,
+            toolChoice: opts.tools ? 'auto' : 'none',
             agentName: 'pilot',
-            stopWhen: opts.task ? () => opts.task.hasFinished : undefined,
+            stopWhen: () => opts.task.hasFinished,
             telemetry: { functionId },
         });
         const text = result?.response?.text || '';
@@ -525,14 +523,10 @@ export class Pilot {
     `;
     }
     getExperienceToc() {
-        if (!this.experienceTracker)
-            return '';
-        const state = this.explorer.getStateManager().getCurrentState();
+        const state = this.stateManager.getCurrentState();
         if (!state)
             return '';
-        const actionResult = ActionResult.fromState(state);
-        const toc = this.experienceTracker.getExperienceTableOfContents(actionResult);
-        return renderExperienceToc(toc);
+        return this.stateManager.getExperienceTracker().renderExperienceTocFor(ActionResult.fromState(state));
     }
     pickPlanningTools() {
         const { see, context, verify, research, getVisitedStates, xpathCheck, learnExperience, askUser } = this.agentTools ?? {};
@@ -600,7 +594,7 @@ export class Pilot {
     async checkDataAvailability(task, requestedData, fishermanReason) {
         if (!this.provider.hasVision())
             return null;
-        const screenshotState = await this.explorer.capturePageWithScreenshot().catch(() => null);
+        const screenshotState = await this.explorer.capture({ screenshot: true }).catch(() => null);
         if (!screenshotState?.screenshot)
             return null;
         const question = dedent `
@@ -646,8 +640,8 @@ export class Pilot {
         else {
             lines.push('modal: none');
         }
-        if (this.explorer.hasOtherTabs()) {
-            const tabs = this.explorer.getOtherTabsInfo();
+        const tabs = this.stateManager.otherTabs;
+        if (tabs.length > 0) {
             lines.push(`other tabs: ${tabs.length} (${tabs.map((t) => `${t.url} - ${t.title}`).join(', ')})`);
         }
         else {
@@ -669,7 +663,7 @@ export class Pilot {
         else {
             lines.push('console errors: none');
         }
-        const failedRequests = this.explorer.getRequestStore()?.getFailedRequests() ?? [];
+        const failedRequests = this.requestStore.getFailedRequests();
         if (failedRequests.length > 0) {
             const sample = failedRequests
                 .slice(-5)
@@ -744,7 +738,7 @@ export class Pilot {
     }
     formatSessionLog(testerConversation) {
         const executions = testerConversation.getToolExecutions().filter((t) => !META_TOOLS.includes(t.toolName));
-        const stateHistory = this.explorer.getStateManager().getStateHistory();
+        const stateHistory = this.stateManager.getStateHistory();
         const initialUrl = stateHistory[0]?.toState?.url || '';
         let currentUrl = initialUrl;
         const groups = new Map();
@@ -886,16 +880,6 @@ export class Pilot {
         return dedent `
       You are Pilot — a supervisor that detects problems and intervenes only when needed.
 
-      SCENARIO: ${task.scenario}
-      START URL: ${initialState.url}
-      PAGE: ${initialState.title || ''} | ${initialState.h1 || ''}
-
-      EXPECTED RESULTS:
-      ${task.expected.map((e) => `- ${e}`).join('\n')}
-
-      PLANNED STEPS:
-      ${stepsText}
-
       Your job: plan, review new pages, detect stuck patterns, suggest concrete next steps. Track which
       expectations are checked. When things go well, encourage briefly and let Tester continue. The current
       page is usually richer than the page summary lists — prefer exploring it before navigating away.
@@ -957,14 +941,24 @@ export class Pilot {
 
       ${dataProtectionRules}
 
-      Describe WHAT to create, not what exists. RIGHT: precondition("1 test"). WRONG:
-      precondition("1 test suite named Updated Suite with existing tests"). Keep descriptions short.
+      Describe WHAT to create, not what exists. RIGHT: precondition("1 post"). WRONG:
+      precondition("1 post named Updated Post with existing comments"). Keep descriptions short.
 
       Response format:
       PROGRESS: <1 sentence assessment>
       NEXT: <specific actionable instruction for Tester>
 
       Keep user-facing reasons concise: one short sentence, maximum 120 characters, evidence only, no repeated verdict wording.
+
+      SCENARIO: ${task.scenario}
+      START URL: ${initialState.url}
+      PAGE: ${initialState.title || ''} | ${initialState.h1 || ''}
+
+      EXPECTED RESULTS:
+      ${task.expected.map((e) => `- ${e}`).join('\n')}
+
+      PLANNED STEPS:
+      ${stepsText}
     `;
     }
 }

package/dist/src/ai/planner.d.ts

@@ -1,14 +1,14 @@
 import { ActionResult } from '../action-result.js';
-import { ExperienceTracker } from '../experience-tracker.js';
+import type { ExperienceTracker } from '../experience-tracker.js';
 import type Explorer from '../explorer.js';
 import type { StateManager } from '../state-manager.js';
 import { Suite } from '../suite.js';
 import { Plan, Test } from '../test-plan.js';
-import type { Agent } from './agent.js';
+import type { Agent, AgentDeps } from './agent.js';
 import { Conversation } from './conversation.js';
 import type { Fisherman } from './fisherman.js';
 import type { Provider } from './provider.js';
-import { Researcher } from './researcher.js';
+import { type Researcher } from './researcher.js';
 declare const PlannerBase: {
     new (...args: any[]): {
         currentPlan: Plan | null;
@@ -58,7 +58,7 @@ export declare class Planner extends PlannerBase implements Agent {
     lastSuite: Suite | null;
     researcher: Researcher;
     fisherman: Fisherman | null;
-    constructor(explorer: Explorer, provider: Provider);
+    constructor(deps: AgentDeps, researcher: Researcher);
     setFisherman(fisherman: Fisherman): void;
     get sectionOrder(): string[];
     getDefaultStartUrl(state: {
@@ -75,5 +75,6 @@ export declare class Planner extends PlannerBase implements Agent {
     cleanExperienceFlows(text: string): string | null;
     buildApproach(style?: string): string;
     buildConversation(state: ActionResult, style?: string, parentPlan?: Plan, feature?: string): Promise<Conversation>;
+    getTasksMessage(): string;
 }
 export {};

package/dist/src/ai/planner.js

@@ -3,7 +3,6 @@ import { z } from 'zod';
 import { ActionResult } from "../action-result.js";
 import { setActivity } from "../activity.js";
 import { ConfigParser } from "../config.js";
-import { ExperienceTracker } from "../experience-tracker.js";
 import { Observability } from "../observability.js";
 import { Stats } from "../stats.js";
 import { Suite } from "../suite.js";
@@ -16,7 +15,7 @@ import { Conversation } from "./conversation.js";
 import { WithSessionDedup } from "./planner/session-dedup.js";
 import { getActiveStyle } from "./planner/styles.js";
 import { WithSubPages, getPlannedByStateHash, getRegisteredPlan, registerPlan } from "./planner/subpages.js";
-import { POSSIBLE_SECTIONS, Researcher } from "./researcher.js";
+import { POSSIBLE_SECTIONS } from "./researcher.js";
 import { findSimilarStateHash } from "./researcher/cache.js";
 import { hasFocusedSection } from "./researcher/focus.js";
 import { capabilityGroundingRule, dataProtectionRules, fileUploadRule } from "./rules.js";
@@ -50,13 +49,13 @@ export class Planner extends PlannerBase {
     lastSuite = null;
     researcher;
     fisherman = null;
-    constructor(explorer, provider) {
+    constructor(deps, researcher) {
         super();
-        this.explorer = explorer;
-        this.provider = provider;
-        this.researcher = new Researcher(explorer, provider);
-        this.stateManager = explorer.getStateManager();
-        this.experienceTracker = new ExperienceTracker();
+        this.explorer = deps.explorer;
+        this.provider = deps.ai;
+        this.researcher = researcher;
+        this.stateManager = deps.stateManager;
+        this.experienceTracker = deps.stateManager.getExperienceTracker();
     }
     setFisherman(fisherman) {
         this.fisherman = fisherman;
@@ -81,7 +80,7 @@ export class Planner extends PlannerBase {
     You are ISTQB certified senior manual QA planning exploratory testing session of a web application.
     Each test scenario must complete a meaningful user workflow — not just open a UI element and verify it exists.
     Bad: "Open the Help panel" — just opens something.
-    Good: "Create a new suite and verify it appears in the list" — completes a business action.
+    Good: "Create a new post and verify it appears in the list" — completes a business action.
     </role>
     <task>
       List possible testing scenarios for the web page.
@@ -149,6 +148,7 @@ export class Planner extends PlannerBase {
         const tags = ['planner'];
         if (style)
             tags.push(style);
+        const existingTestScenarios = this.getExistingTestFileScenarios(state.url);
         const result = await Observability.run(`planner: ${state.url}`, { tags, sessionId: state.url }, async () => {
             const actionResult = ActionResult.fromState(state);
             const conversation = await this.buildConversation(actionResult, style, parentPlan, feature);
@@ -182,7 +182,6 @@ export class Planner extends PlannerBase {
             if (parentPlan)
                 this.currentPlan.parentPlan = parentPlan;
             const allPreviousScenarios = this.getPreviousSessionScenarios();
-            const existingTestScenarios = this.getExistingTestFileScenarios(state.url);
             for (const s of existingTestScenarios)
                 allPreviousScenarios.add(s);
             for (const t of tests) {
@@ -232,6 +231,7 @@ export class Planner extends PlannerBase {
         return added;
     }
     getExistingTestFileScenarios(currentUrl) {
+        this.lastSuite = null;
         if (!currentUrl)
             return new Set();
         try {
@@ -244,27 +244,37 @@ export class Planner extends PlannerBase {
         }
     }
     cleanExperienceFlows(text) {
-        const seenTitles = new Set();
         let result = text;
-        for (const section of [...mdq(text).query('section2').each(), ...mdq(text).query('section3').each()]) {
-            const heading = section.query('heading').text().trim();
-            const body = mdq(section.text())
-                .query('heading')
-                .replace('')
-                .replace(/^---\s*$/gm, '')
-                .trim();
-            if (!body || seenTitles.has(heading)) {
-                result = result.replace(section.text(), '');
-                continue;
-            }
-            seenTitles.add(heading);
-            const blockquotes = section.query('blockquote').each();
-            if (blockquotes.length <= 10)
-                continue;
-            for (const bq of blockquotes.slice(10)) {
-                result = result.replace(bq.text(), '');
-            }
-            result = result.replace(section.text().trim(), `${section.text().trim()}\n> ... and ${blockquotes.length - 10} more discoveries`);
+        const seenTitles = new Set();
+        for (const selector of ['section2', 'section3']) {
+            result = mdq(result)
+                .query(selector)
+                .replaceEach((section) => {
+                const heading = section.query('heading').text().trim();
+                const withoutHeadings = mdq(section.text()).query('heading').replace('');
+                const body = mdq(withoutHeadings).query('hr').replace('').trim();
+                if (body && !seenTitles.has(heading)) {
+                    seenTitles.add(heading);
+                    return section.text();
+                }
+                return '';
+            });
+        }
+        const trimmedTitles = new Set();
+        for (const selector of ['section2', 'section3']) {
+            result = mdq(result)
+                .query(selector)
+                .replaceEach((section) => {
+                const heading = section.query('heading').text().trim();
+                if (trimmedTitles.has(heading))
+                    return section.text();
+                const count = section.query('blockquote').count();
+                if (count <= 10)
+                    return section.text();
+                const kept = mdq(section.text()).query('blockquote[10:]').replace('');
+                trimmedTitles.add(heading);
+                return `${kept.trimEnd()}\n> ... and ${count - 10} more discoveries\n`;
+            });
         }
         return result.trim() || null;
     }
@@ -298,6 +308,7 @@ export class Planner extends PlannerBase {
       If there are subpages (pages with same URL path) plan testing of those subpages as well
       If you plan to test CRUD operations, plan them in correct order: create, read, update.
       Do not invent specific route names, success messages, validation texts, badge counts, or welcome messages unless they are visible in research, visited pages, or prior observed flows.
+      When validation placement or wording was not observed, require feedback associated with the invalid input without inventing a specific location or message.
       If exact wording is unknown, describe the expected result generically, for example "an authentication error is shown" or "the user stays on the login page" instead of guessing the literal text.
       If exact redirect destination is unknown, describe the destination by visible page identity, for example "the dashboard page opens" or "the current workspace home page opens" instead of inventing a URL slug.
       Only propose scenarios whose prerequisites are evident from page research, visited pages, or API data preparation context.
@@ -323,31 +334,26 @@ export class Planner extends PlannerBase {
       ${dataProtectionRules}
       ${fileUploadRule}
       </rules>
-
-      ${this.buildApproach(style)}
-
-      <context>
-      URL: ${state.url || 'Unknown'}
-      Title: ${state.title || 'Unknown'}
-      </context>
     `;
         conversation.addUserText(planningPrompt);
+        conversation.addUserText(this.getTasksMessage());
         const currentState = this.stateManager.getCurrentState();
         const research = await this.researcher.research(currentState || state, {
             deep: true,
         });
         let plannerResearch = mdq(research).query('code').replace('');
-        for (const table of mdq(plannerResearch).query('table').each()) {
-            const rawTable = table.text();
+        plannerResearch = mdq(plannerResearch)
+            .query('table')
+            .replaceEach((table) => {
             const rows = table.toJson();
             if (rows.length === 0 || !rows[0].Element)
-                continue;
+                return table.text();
             const elementWithType = rows.map((r) => ({
                 Element: r.Element,
                 Type: r.Type || '',
             }));
-            plannerResearch = plannerResearch.replace(rawTable, jsonToTable(elementWithType, ['Element', 'Type']));
-        }
+            return jsonToTable(elementWithType, ['Element', 'Type']);
+        });
         const hasFocusedOverlay = hasFocusedSection(plannerResearch);
         const focusNote = hasFocusedOverlay ? "IMPORTANT: One section is marked as **Focused** — this is the user's current focus area. Concentrate testing on the Focused section FIRST — test all interactions inside it before planning tests for the rest of the page." : '';
         const featureFilter = feature ? `FOCUS FILTER: Only propose scenarios using elements relevant to "${feature}". Ignore all other elements.` : '';
@@ -361,6 +367,14 @@ export class Planner extends PlannerBase {
 
       ${plannerResearch}
       </page_research>
+    `);
+        conversation.addUserText(dedent `
+      ${this.buildApproach(style)}
+
+      <context>
+      URL: ${state.url || 'Unknown'}
+      Title: ${state.title || 'Unknown'}
+      </context>
     `);
         if (this.fisherman) {
             await this.fisherman.ensureReady(state.url);
@@ -505,8 +519,10 @@ export class Planner extends PlannerBase {
         </parent_page_context>
       `);
         }
-        const hasCurrentPlan = !!this.currentPlan;
-        const tasksMessage = dedent `
+        return conversation;
+    }
+    getTasksMessage() {
+        return dedent `
     <task>
       Provide testing scenarios as structured data with the following requirements:
       1. Create a short, descriptive plan name that summarizes what will be tested (e.g., "User Authentication Testing", "Product Catalog Navigation", "Form Validation Tests")
@@ -525,17 +541,17 @@ export class Planner extends PlannerBase {
 
          STEPS - actionable commands:
          - Each step should be a specific action to perform
-         - Good: "Click Create Suite button", "Enter 'My New Suite' as suite name", "Click Save button"
-         - Good: "Click Star icon on 'Template API Testing' suite"
+         - Good: "Click Create Post button", "Enter 'My New Post' as post title", "Click Save button"
+         - Good: "Click Star icon on 'Quarterly Report' item"
          - Bad: "Open the dropdown menu" (opening is not a goal, it's a means)
          - Bad: "Verify modal appears" (verification belongs in expected outcomes, not steps)
          - Steps should form a complete workflow, not stop at opening a UI element
          - NEVER split a workflow across two tests. One test = one complete action + its verification
 
          EXPECTED OUTCOMES - verifiable results:
-         - Good: "New suite 'My New Suite' appears in the suite list"
-         - Good: "Suite appears under Starred filter tab"
-         - Good: "Success message 'Suite created' is displayed"
+         - Good: "New post 'My New Post' appears in the posts list"
+         - Good: "Item appears under Starred filter tab"
+         - Good: "Success message 'Post created' is displayed"
          - Good when wording is unknown: "An authentication error is displayed"
          - Good when route is unknown: "The workspace home page is displayed"
          - Bad: "Modal is displayed" (just verifying existence, no business value)
@@ -552,10 +568,8 @@ export class Planner extends PlannerBase {
          - Do not wrap text in ** or * quotes, ( or ) brackets.
          - Avoid using emojis or special characters.
       7. Only tests that can be tested from web UI should be proposed.
-      ${hasCurrentPlan ? '8. CRITICAL: Return ONLY NEW scenarios not in the existing tests list. Return empty array if no new tests needed.' : `8. At least ${this.MIN_TASKS} tests should be proposed.`}
+      8. For a fresh plan propose at least ${this.MIN_TASKS} tests; when expanding an existing plan return ONLY new scenarios not in the tested list, and an empty array if none remain.
     </task>
     `;
-        conversation.addUserText(tasksMessage);
-        return conversation;
     }
 }

package/dist/src/ai/provider.d.ts

@@ -12,7 +12,6 @@ export declare class Provider {
     telemetryEnabled: boolean;
     otelSdk: NodeSDK | null;
     defaultRetryOptions: RetryOptions;
-    static readonly CONTEXT_LENGTH_PATTERNS: string[];
     lastConversation: Conversation | null;
     constructor(config: AIConfig);
     getModelName(model: any): string;
@@ -26,6 +25,10 @@ export declare class Provider {
     getRetryOptions(options?: any): RetryOptions;
     mergeProviderOptions(config: Record<string, any>, agentName?: string): Record<string, any>;
     finalizeConfig(config: Record<string, any>, options: any, telemetry: any): void;
+    buildGenerateConfig(defaults: Record<string, any>, overrides: Record<string, any>, options: any): Record<string, any>;
+    recordUsage(agentName: string, modelName: string, usage: any): void;
+    raceWithIdleTimeout<T>(fn: (signal: AbortSignal) => Promise<T>, timeoutMs: number): Promise<T>;
+    recoverFromContextLength(error: any, messages: ModelMessage[], options: any, retry: (messages: ModelMessage[], options: any) => Promise<any>): Promise<any>;
     initLangfuse(): void;
     getTelemetry(options: any): any;
     startConversation(systemMessage: string, agentName?: string, model?: any): Conversation;