explorbot 0.1.24 → 0.1.26

package/dist/src/components/LogPane.js

@@ -10,7 +10,6 @@ const LogPane = React.memo(({ verboseMode }) => {
     const [logs, setLogs] = useState([]);
     const pendingLogsRef = React.useRef([]);
     const flushTimeoutRef = React.useRef(null);
-    const MAX_MULTILINE_LINES = 16;
     const MAX_STEP_LINES = 8;
     const MAX_SUBSTEP_LINES = 6;
     const formatCollapsedContent = useCallback((lines, collapsedCount, label) => {
@@ -88,6 +87,8 @@ const LogPane = React.memo(({ verboseMode }) => {
                 return { color: 'yellow' };
             case 'debug':
                 return { color: 'gray', dimColor: true };
+            case 'operation':
+                return { color: 'gray', dimColor: true };
             case 'substep':
                 return { color: 'gray', dimColor: true };
             case 'step':
@@ -116,7 +117,8 @@ const LogPane = React.memo(({ verboseMode }) => {
             const cleaned = stripAnsi(dedent(log.content));
             const parsed = parseMarkdownToTerminal(cleaned);
             const lines = parsed.split('\n');
-            const truncated = lines.length > MAX_MULTILINE_LINES ? `${lines.slice(0, MAX_MULTILINE_LINES).join('\n')}\n... (${lines.length - MAX_MULTILINE_LINES} more lines)` : parsed;
+            const maxLines = log.maxLines || 16;
+            const truncated = lines.length > maxLines ? `${lines.slice(0, maxLines).join('\n')}\n... (${lines.length - maxLines} more lines)` : parsed;
             return (React.createElement(Box, { key: index, borderStyle: "classic", borderLeft: false, borderRight: false, marginY: 1, padding: 1, borderColor: "dim", overflow: "hidden" },
                 React.createElement(Text, { color: "gray", dimColor: true }, truncated)));
         }
@@ -127,6 +129,7 @@ const LogPane = React.memo(({ verboseMode }) => {
                 type: 'multiline',
                 content: `HTML Content:\n\n${markdown}`,
                 timestamp: log.timestamp,
+                maxLines: 10,
             };
             return renderLogEntry(multilineLog, index);
         }
@@ -134,6 +137,9 @@ const LogPane = React.memo(({ verboseMode }) => {
         if (log.type === 'substep') {
             return (React.createElement(Box, { key: index, marginLeft: 2, flexDirection: "column" }, lines.map((line, lineIndex) => (React.createElement(Text, { key: `${index}-${lineIndex}`, ...styles }, lineIndex === 0 ? `> ${line}` : `   ${line}`)))));
         }
+        if (log.type === 'operation') {
+            return (React.createElement(Box, { key: index, marginLeft: 2, flexDirection: "column" }, lines.map((line, lineIndex) => (React.createElement(Text, { key: `${index}-${lineIndex}`, ...styles }, lineIndex === 0 ? `· ${line}` : `  ${line}`)))));
+        }
         if (log.type === 'step') {
             return (React.createElement(Box, { key: index, flexDirection: "column", paddingLeft: 2 }, lines.map((line, lineIndex) => (React.createElement(Text, { key: `${index}-${lineIndex}`, ...styles }, line)))));
         }
@@ -145,7 +151,7 @@ const LogPane = React.memo(({ verboseMode }) => {
             icon && React.createElement(Text, { ...styles }, icon),
             React.createElement(Box, { flexDirection: "column" }, lines.map((line, lineIndex) => (React.createElement(Text, { key: `${index}-${lineIndex}`, ...styles }, line))))));
     };
-    const maxLogs = 100;
+    const maxLogs = 80;
     const visibleLogs = logs.length > maxLogs ? logs.slice(-maxLogs) : logs;
     return React.createElement(Box, { flexDirection: "column" }, visibleLogs.map((log, index) => renderLogEntry(log, index)).filter(Boolean));
 });

package/dist/src/experience-tracker.js

@@ -164,7 +164,7 @@ export class ExperienceTracker {
         const newEntry = generateActionContent(title, filteredCode, action.explanation);
         const updatedContent = `${newEntry}\n\n${content}`;
         this.writeExperienceFile(stateHash, updatedContent, data);
-        tag('substep').log(` Added ACTION to: ${stateHash}.md`);
+        tag('operation').log(`Added ACTION to: ${stateHash}.md`);
     }
     writeFlow(state, body, relatedUrls) {
         if (this.disabled || this.isWritingDisabled(state))
@@ -190,7 +190,7 @@ export class ExperienceTracker {
         }
         const updatedContent = `${body}\n${content}`;
         this.writeExperienceFile(stateHash, updatedContent, data);
-        tag('substep').log(`Added FLOW to: ${stateHash}.md`);
+        tag('operation').log(`Added FLOW to: ${stateHash}.md`);
     }
     getAllExperience() {
         const allFiles = [];

package/dist/src/explorbot.js

@@ -422,7 +422,7 @@ export class ExplorBot {
                 this.lastReportedTestCount = tests.length;
                 return;
             }
-            tag('multiline').log(markdown);
+            tag('multiline').log(markdown, { maxLines: 22 });
             const filePath = this.agentSessionAnalyst().writeReport(markdown);
             tag('info').log(`Session report saved: ${relativeToCwd(filePath)}`);
             const reporter = this.explorer?.getReporter();

package/dist/src/reporter.js

@@ -82,10 +82,12 @@ export class Reporter {
             return;
         }
         try {
-            this.client = new Client({ apiKey: process.env.TESTOMATIO || '', title: this.buildTitle() });
-            const timeoutMs = Number(process.env.TESTOMATIO_TIMEOUT_MS || '15000');
-            const timeoutPromise = new Promise((resolve) => setTimeout(() => resolve('timeout'), timeoutMs));
-            const result = await Promise.race([this.client.createRun({ configuration: { exploratory: true } }).then(() => 'success'), timeoutPromise]);
+            const result = await withQuietReporterLogs(async () => {
+                this.client = new Client({ apiKey: process.env.TESTOMATIO || '', title: this.buildTitle() });
+                const timeoutMs = Number(process.env.TESTOMATIO_TIMEOUT_MS || '15000');
+                const timeoutPromise = new Promise((resolve) => setTimeout(() => resolve('timeout'), timeoutMs));
+                return await Promise.race([this.client.createRun({ configuration: { exploratory: true } }).then(() => 'success'), timeoutPromise]);
+            });
             if (result === 'timeout') {
                 debugLog('Reporter run creation timed out');
                 return;
@@ -252,7 +254,9 @@ export class Reporter {
             return;
         }
         try {
-            await this.client.updateRunStatus('finished');
+            await withQuietReporterLogs(async () => {
+                await this.client.updateRunStatus('finished');
+            });
             this.isRunStarted = false;
             debugLog('Testomat.io run finished');
         }
@@ -299,3 +303,18 @@ export class Reporter {
         return;
     }
 }
+async function withQuietReporterLogs(fn) {
+    const previousLevel = process.env.TESTOMATIO_LOG_LEVEL;
+    process.env.TESTOMATIO_LOG_LEVEL = 'ERROR';
+    try {
+        return await fn();
+    }
+    finally {
+        if (previousLevel === undefined) {
+            process.env.TESTOMATIO_LOG_LEVEL = undefined;
+        }
+        else {
+            process.env.TESTOMATIO_LOG_LEVEL = previousLevel;
+        }
+    }
+}

package/dist/src/utils/log-filters.js

@@ -0,0 +1,27 @@
+export class RecentStepFilter {
+    ttlMs;
+    recentStepKeys = new Map();
+    constructor(ttlMs = 15000) {
+        this.ttlMs = ttlMs;
+    }
+    shouldSuppress(content, now = Date.now()) {
+        const key = normalizeStepCommand(content);
+        if (!key)
+            return false;
+        for (const [existingKey, timestamp] of this.recentStepKeys) {
+            if (now - timestamp > this.ttlMs) {
+                this.recentStepKeys.delete(existingKey);
+            }
+        }
+        if (this.recentStepKeys.has(key))
+            return true;
+        this.recentStepKeys.set(key, now);
+        return false;
+    }
+}
+function normalizeStepCommand(content) {
+    const normalized = content.replace(/\s+/g, ' ').trim();
+    if (!normalized.startsWith('I.'))
+        return null;
+    return normalized.toLowerCase();
+}

package/dist/src/utils/logger.js

@@ -7,6 +7,7 @@ import dedent from 'dedent';
 import stripAnsi from 'strip-ansi';
 import { ConfigParser } from '../config.js';
 import { Observability } from "../observability.js";
+import { RecentStepFilter } from "./log-filters.js";
 import { parseMarkdownToTerminal } from "./markdown-terminal.js";
 class DebugFilter {
     patterns = [];
@@ -51,6 +52,7 @@ const debugFilter = new DebugFilter();
 class ConsoleDestination {
     verboseMode = false;
     forceEnabled = false;
+    recentSteps = new RecentStepFilter();
     isEnabled() {
         return this.forceEnabled || !process.env.INK_RUNNING;
     }
@@ -65,6 +67,10 @@ class ConsoleDestination {
             return;
         if (entry.type === 'html')
             return;
+        if (entry.type === 'operation' && !this.verboseMode)
+            return;
+        if (entry.type === 'step' && !this.verboseMode && this.recentSteps.shouldSuppress(entry.content))
+            return;
         let content = entry.content;
         if (entry.type === 'multiline') {
             const cleaned = stripAnsi(dedent(entry.content));
@@ -84,6 +90,9 @@ class ConsoleDestination {
         else if (entry.type === 'step') {
             content = chalk.gray(`   ${content}`);
         }
+        else if (entry.type === 'operation') {
+            content = chalk.gray(`   · ${content}`);
+        }
         else if (entry.type === 'substep') {
             content = chalk.gray(`   > ${content}`);
         }
@@ -213,6 +222,8 @@ class ReactDestination {
         return this.debugMode;
     }
     shouldWrite(entry) {
+        if (entry.type === 'operation' && !this.debugMode)
+            return false;
         if (entry.type !== 'debug')
             return true;
         if (this.debugMode)
@@ -269,7 +280,7 @@ class CaptainDestination {
     }
     stopCapture() {
         this.capturing = false;
-        const logs = this.entries.filter((e) => e.type !== 'debug' && e.type !== 'html' && e.type !== 'multiline').map((e) => `[${e.type}] ${e.content}`);
+        const logs = this.entries.filter((e) => e.type !== 'debug' && e.type !== 'html' && e.type !== 'multiline' && e.type !== 'operation').map((e) => `[${e.type}] ${e.content}`);
         this.entries = [];
         return logs;
     }
@@ -377,6 +388,7 @@ class Logger {
             }
             return;
         }
+        const options = this.extractLogOptions(type, args);
         let content = this.processArgs(args);
         if (type === 'step' && args[0]?.toCode) {
             content = args[0].toCode();
@@ -386,6 +398,7 @@ class Logger {
             content,
             timestamp: new Date(),
             originalArgs: args,
+            maxLines: options?.maxLines,
         };
         if (this.file.isEnabled())
             this.file.write(entry);
@@ -421,6 +434,20 @@ class Logger {
     multiline(...args) {
         this.log('multiline', ...args);
     }
+    extractLogOptions(type, args) {
+        if (type !== 'multiline')
+            return null;
+        const last = args[args.length - 1];
+        if (!last || typeof last !== 'object' || Array.isArray(last))
+            return null;
+        if (!('maxLines' in last))
+            return null;
+        args.pop();
+        const maxLines = Number(last.maxLines);
+        if (!Number.isFinite(maxLines) || maxLines <= 0)
+            return null;
+        return { maxLines };
+    }
 }
 const logger = Logger.getInstance();
 let stepSpanParent = null;

package/dist/src/utils/next-steps.js

@@ -27,11 +27,5 @@ export function printNextSteps(sections) {
         }
         blocks.push(lines.join('\n'));
     }
-    for (let i = 0; i < blocks.length; i++) {
-        if (i > 0)
-            tag('info').log('');
-        for (const line of blocks[i].split('\n')) {
-            tag('info').log(line);
-        }
-    }
+    tag('multiline').log(blocks.join('\n\n'), { maxLines: 18 });
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "explorbot",
-  "version": "0.1.24",
+  "version": "0.1.26",
   "description": "CLI app built with React Ink, CodeceptJS, and Playwright",
   "license": "Elastic-2.0",
   "type": "module",

package/src/ai/historian/codeceptjs.ts

@@ -102,7 +102,7 @@ export function WithCodeceptJS<T extends Constructor>(Base: T) {
       writeFileSync(filePath, lines.join('\n'));
       this.savedFiles.add(filePath);
 
-      tag('substep').log(`Saved plan tests to: ${relativeToCwd(filePath)}`);
+      tag('operation').log(`Saved plan tests to: ${relativeToCwd(filePath)}`);
       return filePath;
     }
 

package/src/ai/historian/experience.ts

@@ -56,7 +56,7 @@ export function WithExperience<T extends Constructor>(Base: T) {
 
       await this.stopScreencast();
 
-      tag('substep').log(`Historian saved session for: ${task.description}`);
+      tag('operation').log(`Historian saved session for: ${task.description}`);
     }
 
     private async reportSession(test: Test, steps: SessionStep[]): Promise<void> {

package/src/ai/historian/playwright.ts

@@ -140,7 +140,7 @@ export function WithPlaywright<T extends Constructor>(Base: T) {
       writeFileSync(filePath, lines.join('\n'));
       this.savedFiles.add(filePath);
 
-      tag('substep').log(`Saved plan tests to: ${relativeToCwd(filePath)}`);
+      tag('operation').log(`Saved plan tests to: ${relativeToCwd(filePath)}`);
       return filePath;
     }
 

package/src/ai/historian/screencast.ts

@@ -92,7 +92,7 @@ export function WithScreencast<T extends Constructor>(Base: T) {
         this.screencastTask = test?._explorbotTest || null;
         this.screencastLastChapter = null;
       } catch (err) {
-        tag('substep').log(`Screencast start failed: ${(err as Error).message}`);
+        tag('operation').log(`Screencast start failed: ${(err as Error).message}`);
       }
     }
 
@@ -116,7 +116,7 @@ export function WithScreencast<T extends Constructor>(Base: T) {
       try {
         await this.screencastPage.screencast.stop();
       } catch (err) {
-        tag('substep').log(`Screencast stop failed: ${(err as Error).message}`);
+        tag('operation').log(`Screencast stop failed: ${(err as Error).message}`);
       }
       this.screencastActive = false;
       this.screencastPage = null;
@@ -126,7 +126,7 @@ export function WithScreencast<T extends Constructor>(Base: T) {
       if (path) {
         this.savedFiles.add(path);
         task?.addArtifact?.(path);
-        tag('substep').log(`Saved screencast: ${relativeToCwd(path)}`);
+        tag('operation').log(`Saved screencast: ${relativeToCwd(path)}`);
       }
     }
   };

package/src/ai/historian.ts

@@ -62,6 +62,6 @@ export class Historian extends HistorianBase {
 
     writeFileSync(filePath, content);
     this.savedFiles.add(filePath);
-    tag('substep').log(`Updated test file with healed steps: ${relativeToCwd(filePath)}`);
+    tag('operation').log(`Updated test file with healed steps: ${relativeToCwd(filePath)}`);
   }
 }

package/src/ai/navigator.ts

@@ -206,7 +206,7 @@ class Navigator implements Agent {
     if (!actionResult.isInsideIframe) {
       const successful = this.experienceTracker.getSuccessfulExperience(actionResult);
       if (successful.length > 0) {
-        tag('substep').log(`Found ${successful.length} experience ${pluralize(successful.length, 'file')} for: ${actionResult.url}`);
+        tag('operation').log(`Found ${successful.length} experience ${pluralize(successful.length, 'file')} for: ${actionResult.url}`);
         experience = `<experience>\nPast successful recipes recorded from prior runs for this page. Prefer these solutions first if they match the goal.\n\n${successful.join('\n\n')}\n</experience>`;
       }
     }
@@ -307,7 +307,7 @@ class Navigator implements Agent {
             stop();
             return;
           }
-          tag('substep').log('Feeding failures back to AI for a new batch...');
+          tag('operation').log('Feeding failures back to AI for a new batch...');
           let contextMsg = 'Previous solutions did not work. Analyze the failures and try DIFFERENT strategies (not syntactic variants of the same locator).\n\n';
           if (batchFailures.length > 0) {
             const lines = batchFailures
@@ -633,7 +633,7 @@ class Navigator implements Agent {
 
     const cachedVerification = actionResult.getVerification(message);
     if (cachedVerification !== null) {
-      tag('substep').log(`Reusing cached verification: ${cachedVerification ? 'PASS' : 'FAIL'}`);
+      tag('operation').log(`Reusing cached verification: ${cachedVerification ? 'PASS' : 'FAIL'}`);
       return { verified: cachedVerification, successfulCodes: [], assertionSteps: [], totalAttempted: 0 };
     }
 
@@ -654,7 +654,7 @@ class Navigator implements Agent {
       const toc = this.experienceTracker.getExperienceTableOfContents(actionResult);
       if (toc.length > 0) {
         const totalSections = toc.reduce((sum, entry) => sum + entry.sections.length, 0);
-        tag('substep').log(`Found ${toc.length} experience ${pluralize(toc.length, 'file')} (${totalSections} sections) for: ${actionResult.url}`);
+        tag('operation').log(`Found ${toc.length} experience ${pluralize(toc.length, 'file')} (${totalSections} sections) for: ${actionResult.url}`);
         experience = renderExperienceToc(toc);
       }
     }

package/src/ai/pilot.ts

@@ -104,7 +104,7 @@ export class Pilot implements Agent {
 
     const schema = z.object({
       decision: z.enum(['pass', 'fail', 'continue', 'skipped']).describe('pass = test succeeded, fail = test failed, continue = tester should keep going, skipped = scenario is irrelevant OR systematic execution failures prevented testing'),
-      reason: z.string().describe('What happened and why (1-2 sentences). Do NOT repeat the decision status (e.g. "scenario goal achieved/not achieved") — just explain the evidence. For continue: explain why rejected and suggest alternatives.'),
+      reason: z.string().describe('Concise user-facing reason, maximum 1 short sentence and 120 characters. Do NOT repeat the decision status; explain only the evidence. For continue: explain why rejected and suggest alternatives.'),
       guidance: z.string().nullable().describe('Required for "continue": specific actionable instruction for the tester — what exactly to verify, retry differently, or complete next. Be concrete.'),
       requestVerification: z
         .string()
@@ -177,7 +177,7 @@ export class Pilot implements Agent {
         }
       }
 
-      tag('info').log(`Pilot: ${result.decision} — ${result.reason}`);
+      tag('info').log(`Pilot: ${result.decision} - ${result.reason}`);
       task.summary = result.reason;
 
       const verdictState = screenshotState || currentState;
@@ -221,7 +221,7 @@ export class Pilot implements Agent {
 
     const schema = z.object({
       decision: z.enum(['allow', 'fail', 'continue', 'skipped']).describe('allow = reset proceeds, fail = test failed (stop looping), continue = veto reset, tester should act on current page instead, skipped = scenario is irrelevant or cannot be executed'),
-      reason: z.string().describe('What evidence justifies this decision (1-2 sentences). Do not restate the decision.'),
+      reason: z.string().describe('Concise evidence-only reason, maximum 1 short sentence and 120 characters. Do not restate the decision.'),
       guidance: z.string().nullable().describe('Required for "continue": concrete instruction for what the tester should do instead of resetting (e.g. which tool to call, what to verify).'),
     });
 
@@ -388,8 +388,9 @@ export class Pilot implements Agent {
       - "continue": tester hasn't completed the goal; provide concrete guidance (which tool, what to check).
         If a verify() asserted a state that was ALREADY TRUE before the test, it proves nothing — reject.
 
-      reason field: do NOT restate the decision ("scenario goal achieved/not achieved"). State what happened —
-      what was verified, what failed, what evidence was found.
+      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.
     `;
   }
 
@@ -1017,6 +1018,8 @@ export class Pilot implements Agent {
       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.
     `;
   }
 }

package/src/ai/planner.ts

@@ -36,7 +36,7 @@ const TasksSchema = z.object({
         scenario: z.string().describe('A single sentence describing what to test'),
         priority: z.enum(['critical', 'important', 'high', 'normal', 'low']).describe('Priority of the task based on business importance'),
         startUrl: z.string().nullable().describe('Start URL for the test if different from plan URL (only for tests on visited subpages)'),
-        steps: z.array(z.string()).describe('List of steps to perform for this scenario. Each step should be a specific action (e.g., "Click on Login button", "Enter username in email field", "Submit the form"). Keep steps atomic and actionable.'),
+        steps: z.array(z.string()).describe('List of steps to perform for this scenario. Each step should be a specific action (e.g., "Open the form", "Enter required data", "Submit the form"). Keep steps atomic and actionable.'),
         expectedOutcomes: z
           .array(z.string())
           .describe('List of expected outcomes that can be verified. Each outcome should be simple, specific, and easy to check (e.g., "Success message appears", "URL changes to /dashboard", "Form field shows error"). Keep outcomes atomic - do not combine multiple checks into one.'),
@@ -226,9 +226,7 @@ export class Planner extends PlannerBase implements Agent {
       }
     }
 
-    const availableStyles = Object.keys(getStyles()).join(', ');
     tag('success').log(`Planning complete! ${this.currentPlan.tests.length} tests in plan: ${this.currentPlan.title}`);
-    tag('info').log(`Planning style: ${this.lastStyleName} (available: ${availableStyles})`);
 
     if (state.url) registerPlan(state.url, this.currentPlan, feature, state.hash);
 

package/src/ai/provider.ts

@@ -294,8 +294,11 @@ export class Provider {
         }
         throw new ContextLengthError(error.message || error.toString());
       }
-      tag('error').log(error.message || error.toString());
-      throw new AiError(error.message || error.toString());
+      const message = error.message || error.toString();
+      if (message !== 'No response text from AI') {
+        tag('error').log(message);
+      }
+      throw new AiError(message);
     }
   }
 
@@ -376,7 +379,6 @@ export class Provider {
     } catch (error: any) {
       clearActivity();
       if (error?.message?.includes('Tool choice is required')) {
-        tag('warning').log('Model completed without calling a tool, returning empty result');
         return { text: '', toolCalls: [], toolResults: [], response: { messages: [] }, usage: null };
       }
       if (error?.name === 'AbortError') throw error;

package/src/ai/researcher/locators.ts

@@ -80,7 +80,7 @@ export function WithLocators<T extends Constructor>(Base: T) {
         }
       }
 
-      tag('substep').log(`Validated ${locators.length} locators: ${locators.length - broken} valid, ${broken} broken`);
+      tag('operation').log(`Validated ${locators.length} locators: ${locators.length - broken} valid, ${broken} broken`);
     }
 
     async fixBrokenSections(result: ResearchResult, conversation: Conversation): Promise<void> {

package/src/ai/researcher.ts

@@ -151,7 +151,7 @@ export class Researcher extends ResearcherBase implements Agent {
       if (!deep && !force) {
         const similar = await findSimilarResearch(combinedHtml);
         if (similar) {
-          tag('substep').log('Similar research found, reusing cached result');
+          tag('operation').log('Similar research found, reusing cached result');
           if (stateHash) saveResearch(stateHash, similar, combinedHtml);
           tag('multiline').log(formatResearchSummary(similar));
           tag('success').log('Research complete (reused)');
@@ -316,10 +316,10 @@ export class Researcher extends ResearcherBase implements Agent {
 
       tag('multiline').log(formatResearchSummary(result.text, { visionUsed: this.hasScreenshotToAnalyze }));
       tag('success').log('Research complete');
-      if (researchFile) tag('substep').log(`Research file saved to: ${researchFile}`);
+      if (researchFile) tag('operation').log(`Research file saved to: ${researchFile}`);
       if (this.actionResult?.screenshotFile) {
         const screenshotPath = outputPath('states', this.actionResult.screenshotFile);
-        tag('substep').log(`UI screenshot: file://${screenshotPath}`);
+        tag('operation').log(`UI screenshot: file://${screenshotPath}`);
       }
 
       await this.hooksRunner.runAfterHook('researcher', state.url);
@@ -467,7 +467,7 @@ export class Researcher extends ResearcherBase implements Agent {
         .filter((k) => !!k)
         .join('\n\n');
 
-      tag('substep').log(`Found ${knowledgeFiles.length} relevant knowledge ${pluralize(knowledgeFiles.length, 'file')} for: ${this.actionResult.url}`);
+      tag('operation').log(`Found ${knowledgeFiles.length} relevant knowledge ${pluralize(knowledgeFiles.length, 'file')} for: ${this.actionResult.url}`);
       knowledge = `
         <hint>
         Here is relevant knowledge for this page:

package/src/ai/rules.ts

@@ -131,6 +131,12 @@ export const fileUploadRule = dedent`
   </file_upload>
 `;
 
+export const formRequirementsRule = dedent`
+  <form_requirements>
+  Before filling a form that persists data (create/update), read each control's requirements (required, type/format, length, placeholder/aria-describedby hints) from <page_aria> and page HTML — call context() if not visible — and enter values that satisfy them. Search/filter/sort forms that only change the view do not need this.
+  </form_requirements>
+`;
+
 // in rage mode we do not protect from irreversible actions
 export const protectionRule = dedent`
   <important>
@@ -270,6 +276,8 @@ export const actionRule = dedent`
   If locator doesn't work, try CSS or XPath locators.
   If nothing works, use I.clickXY(x, y) as last resort.
 
+  For checkboxes, prefer I.checkOption/I.uncheckOption over I.click.
+
 
   ### I.fillField
 
@@ -356,6 +364,19 @@ export const actionRule = dedent`
     I.selectOption('form select[name=account]', 'Premium');
   </example>
 
+  ### I.checkOption / I.uncheckOption
+
+  Set a checkbox/radio to a definite state — idempotent, never toggles. Use for checkboxes instead of I.click. Run via form(), not click().
+
+  I.checkOption(<locator>, <context>)
+  I.uncheckOption(<locator>, <context>)
+
+  <example>
+    I.checkOption('Subscribe');
+    I.checkOption({ role: 'checkbox', text: 'Agree' });
+    I.uncheckOption('Subscribe', '.preferences');
+  </example>
+
   ### I.attachFile
 
   Attaches a file to a file input element.

package/src/ai/session-analyst.ts

@@ -41,6 +41,9 @@ export class SessionAnalyst implements Agent {
 
       Crucial distinction: "the app misbehaved" vs "the automation could not interact with the app". ONLY the first is a Defect. If the automation gives up before the app responds — timeout, retries exhausted, dead loop / loop detected, could not click or find an element — that is an Execution issue regardless of what the log calls it. Failure inside the automation ≠ failure inside the product.
 
+      The action log is more authoritative than the scenario title. If the actual submitted data, page state, or action sequence does not match the scenario title, classify it as Execution issue and do not list that scenario under What works. Do NOT infer a product Defect or UX issue from behavior caused by incorrect test data or an automation mismatch.
+      Negative test data is valid when it matches a negative scenario. Do not call intentionally invalid input wrong data when the scenario expects rejection or validation feedback.
+
       A solitary failure where adjacent tests on the same feature passed → Execution, not Defect.
 
       ## Severity (defects only)
@@ -76,7 +79,7 @@ export class SessionAnalyst implements Agent {
 
       ## Brevity rules
 
-      - Headline: 2 sentences MAX. About the FEATURE, not the run. No counts, no "N tests", no "this session". Banned words: "exercised", "comprehensive", "notably", "this session", "module", "targeted", "covered creation".
+      - Headline: 2 sentences MAX. About the FEATURE, not the run. No counts, no "N tests", no "this session". Never use these words: "exercised", "comprehensive", "notably", "this session", "module", "targeted", "covered creation".
       - What works: feature name + test refs. NO parentheticals, NO caveats. If there's a caveat, the entry doesn't belong here.
       - Defect title is the BUG ("Search returns non-matching results"), never the scenario name.
       - Reproduce steps are imperative one-liners drawn from the log.

package/src/ai/task-agent.ts

@@ -44,7 +44,7 @@ export abstract class TaskAgent {
       .filter((k) => !!k)
       .join('\n\n');
 
-    tag('substep').log(`Found ${knowledgeFiles.length} relevant knowledge ${pluralize(knowledgeFiles.length, 'file')}`);
+    tag('operation').log(`Found ${knowledgeFiles.length} relevant knowledge ${pluralize(knowledgeFiles.length, 'file')}`);
     return dedent`
       <knowledge>
       Here is relevant knowledge for this page:
@@ -61,7 +61,7 @@ export abstract class TaskAgent {
 
     const totalSections = toc.reduce((sum, entry) => sum + entry.sections.length, 0);
     debugLog(`injecting experience TOC (${toc.length} files, ${totalSections} sections)`);
-    tag('substep').log(`Found ${toc.length} experience ${pluralize(toc.length, 'file')} (${totalSections} sections)`);
+    tag('operation').log(`Found ${toc.length} experience ${pluralize(toc.length, 'file')} (${totalSections} sections)`);
     return renderExperienceToc(toc);
   }
 

package/src/ai/tester.ts

@@ -25,7 +25,7 @@ import { Navigator } from './navigator.ts';
 import type { Pilot } from './pilot.ts';
 import { Provider } from './provider.ts';
 import { Researcher } from './researcher.ts';
-import { actionRule, focusedElementRule, locatorRule, multipleTabsRule, protectionRule, sectionContextRule } from './rules.ts';
+import { actionRule, focusedElementRule, formRequirementsRule, locatorRule, multipleTabsRule, protectionRule, sectionContextRule } from './rules.ts';
 import { TaskAgent } from './task-agent.ts';
 import { createCodeceptJSTools, createSpecialContextTools } from './tools.ts';
 
@@ -773,6 +773,8 @@ export class Tester extends TaskAgent implements Agent {
 
     ${sectionContextRule}
 
+    ${formRequirementsRule}
+
     ${this.provider.getSystemPromptForAgent('tester', this.explorer.getStateManager().getCurrentState()?.url) || ''}
     `;
   }

package/src/ai/tools.ts

@@ -307,6 +307,7 @@ export function createCodeceptJSTools(explorer: Explorer, task: Task) {
 
         Use cases:
         - Typing into input fields (I.fillField, I.type)
+        - Setting checkboxes/radios to a definite state (I.checkOption, I.uncheckOption)
         - Working with iframes (switch context with I.switchTo)
         - Performing multiple form actions in a single batch
         - Complex interactions requiring sequential commands
@@ -314,6 +315,7 @@ export function createCodeceptJSTools(explorer: Explorer, task: Task) {
         Example - filling a form with context (PREFERRED):
         I.fillField('Username', 'John', '.login-form')
         I.selectOption('Country', 'USA', '.address-section')
+        I.checkOption('Agree', '.terms-section')
         I.attachFile('input[type="file"]', 'path/to/file', '.upload-section')
 
         Example - filling a form with ARIA locators:

package/src/commands/context-aria-command.ts

@@ -17,6 +17,6 @@ export class ContextAriaCommand extends BaseCommand {
       throw new Error('No ARIA snapshot available for current page');
     }
 
-    tag('multiline').log(`ARIA Snapshot:\n\n${ariaSnapshot}`);
+    tag('multiline').log(`ARIA Snapshot:\n\n${ariaSnapshot}`, { maxLines: 10 });
   }
 }