@avi770/testteam 2.0.0 → 3.0.0

package/CHANGELOG.md

@@ -5,6 +5,29 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.0.0] - 2026-03-30
+
+### Added
+- **Total Recall integration** — persistent memory for findings history and corrections learning across runs
+- **Agent-MVP integration** — reputation scoring, tier progression, coaching flags for all 70 agents
+- **MAH integration** — Multi-Agent Hub support with 100-agent limit, CustomExecutor, code-execution skills
+- **Execution Bridge** — new `bridge/` module with pluggable executors:
+  - `PlaywrightPool` — reusable browser instance pool (max 3 concurrent)
+  - `ShellExecutor` — sandboxed child process execution with timeout enforcement
+  - `FileScanner` — regex-based source file scanning with glob filtering
+  - `HttpClient` — native fetch wrapper with concurrent request support
+  - `BridgeRouter` — task routing to correct executor, including hybrid multi-step pipelines
+- `--coaching <agentId>` CLI flag to query Agent-MVP coaching flags
+- `--recall <query>` CLI flag to query Total Recall memory
+- `--reputation` CLI flag to display agent leaderboard
+- Startup banner showing version and connection status for all 3 services (TR, Agent-MVP, MAH)
+- Docker Compose stack for full local development (Total Recall, Agent-MVP, PostgreSQL, Redis)
+
+### Changed
+- Version bumped to 3.0.0 (major: learning/reputation system is a fundamental capability change)
+- CLI startup now prints "TestTeam v3.0.0 — Learning Agents" banner
+- Package description updated to reflect learning + reputation capabilities
+
 ## [2.0.0] - 2026-03-30
 
 ### Added
@@ -77,6 +100,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - 595 unit tests
 - TypeScript strict mode throughout with full type safety
 
-[2.0.0]: https://github.com/BAS-More/E2E-Tester/compare/v1.1.0...HEAD
+[3.0.0]: https://github.com/BAS-More/E2E-Tester/compare/v2.0.0...HEAD
+[2.0.0]: https://github.com/BAS-More/E2E-Tester/compare/v1.1.0...v2.0.0
 [1.1.0]: https://github.com/BAS-More/E2E-Tester/compare/v1.0.0...v1.1.0
 [1.0.0]: https://github.com/BAS-More/E2E-Tester/releases/tag/v1.0.0

package/README.md

@@ -18,7 +18,7 @@
 
 ---
 
-Autonomous multi-phase testing pipeline with 70 specialized agents. Runs Alpha through Release with quality gates, self-healing fix loops, and zero manual intervention.
+Autonomous multi-phase testing pipeline with 70 learning agents. Runs Alpha through Release with quality gates, self-healing fix loops, reputation scoring, persistent memory, and zero manual intervention.
 
 ## Quick Start
 
@@ -119,15 +119,30 @@ The init wizard auto-detects your framework, test runner, and scaffolds configur
 | 69 | Type Safety Auditor | Alpha, Beta, UAT | Pro |
 | 70 | Complexity Analyzer | Alpha, Beta, UAT | Pro |
 
+## v3.0.0 — Learning Agents
+
+TestTeam v3.0.0 introduces a learning layer that makes agents smarter over time:
+
+- **Total Recall** — persistent memory stores findings, corrections, and context across runs. Agents recall past issues and avoid false positives.
+- **Agent-MVP** — reputation scoring with tier progression (Rookie through Legend). High-performing agents earn trust; underperformers get coaching flags.
+- **MAH (Multi-Agent Hub)** — orchestration layer supporting up to 100 concurrent agents with CustomExecutor and code-execution skills.
+- **Execution Bridge** — pluggable executor system with PlaywrightPool (browser reuse), ShellExecutor (sandboxed commands), FileScanner (regex source scanning), and HttpClient (concurrent API testing).
+
 ## Architecture
 
 ```
+                  Total Recall (Memory)
+                         |
 Alpha --> [gate] --> Beta --> [gate] --> UAT --> [gate] --> Release
   |                   |                  |
   +------- Fix Loop (Healer + Fixer) ---+
+  |                                      |
+  +---- Execution Bridge (Pool/Shell/HTTP/Scan) ----+
+  |                                                  |
+  +------------ Agent-MVP (Reputation) -------------+
 ```
 
-Each phase dispatches its agents, evaluates findings against blocking/advisory classification, and promotes to the next phase only if the gate passes. The fix loop runs automatically when agents detect issues.
+Each phase dispatches its agents, evaluates findings against blocking/advisory classification, and promotes to the next phase only if the gate passes. The fix loop runs automatically when agents detect issues. Agent performance is tracked by Agent-MVP and findings are persisted in Total Recall.
 
 ## Pricing
 
@@ -176,6 +191,9 @@ Options:
   --dry-run                                Show dispatch plan without executing
   --agent <id>                             Run a single agent by ID
   --config <path>                          Path to config file
+  --recall <query>                         Query Total Recall memory
+  --reputation                             Show agent reputation leaderboard
+  --coaching <agentId>                     Show coaching flags for an agent
   --help                                   Show help
 ```
 

package/agents/15-regression-sentinel.ts

@@ -2,6 +2,7 @@ import { mkdir, readdir, readFile, writeFile, stat } from 'node:fs/promises';
 import { join } from 'path';
 import type { Finding, Phase } from '../core/types';
 import type { ValidatedConfig } from '../core/config';
+import type { TotalRecallClient } from '../clients/total-recall';
 import { BaseAgent } from './base-agent';
 import { compareScreenshots, updateBaseline } from '../helpers/screenshot';
 
@@ -27,8 +28,8 @@ export class RegressionSentinelAgent extends BaseAgent {
   /** Absolute path to the baselines directory (sibling of runDir). */
   private readonly baselinesDir: string;
 
-  constructor(config: ValidatedConfig, phase: Phase, runDir: string) {
-    super(config, phase, runDir);
+  constructor(config: ValidatedConfig, phase: Phase, runDir: string, totalRecall?: TotalRecallClient) {
+    super(config, phase, runDir, totalRecall);
     // Place baselines as a sibling directory to runDir
     const parentDir = join(runDir, '..');
     this.baselinesDir = join(parentDir, BASELINES_DIR_NAME);

package/agents/base-agent.ts

@@ -1,11 +1,13 @@
 import type { AgentResult, Finding, Phase, EvidenceItem } from '../core/types';
 import type { ValidatedConfig } from '../core/config';
+import type { TotalRecallClient } from '../clients/total-recall';
 
 export abstract class BaseAgent {
   constructor(
     protected readonly config: ValidatedConfig,
     protected readonly phase: Phase,
     protected readonly runDir: string,
+    protected readonly totalRecall?: TotalRecallClient,
   ) {}
 
   abstract readonly agentId: number;
@@ -16,6 +18,12 @@ export abstract class BaseAgent {
     try {
       await this.preFlight();
       const findings = await this.execute();
+
+      // Store findings in Total Recall if available (fire-and-forget)
+      if (this.totalRecall && findings.length > 0) {
+        await this.totalRecall.storeFindings(this.agentId, this.agentName, findings).catch(() => {});
+      }
+
       const completedAt = new Date().toISOString();
       return {
         agentId: this.agentId,

package/agents/registry.ts

@@ -1,6 +1,7 @@
 import type { ValidatedConfig } from '../core/config';
 import type { Phase } from '../core/types';
 import type { Orchestrator } from '../core/orchestrator';
+import type { TotalRecallClient } from '../clients/total-recall';
 
 /** License tier type used for agent gating */
 export type LicenseTier = 'free' | 'pro' | 'enterprise';
@@ -197,78 +198,79 @@ export function registerAllAgents(
   config: ValidatedConfig,
   phase: Phase,
   runDir: string,
+  totalRecall?: TotalRecallClient,
 ): void {
   const agents = [
-    new AnalystAgent(config, phase, runDir),
-    new SeedArchitectAgent(config, phase, runDir),
-    new TestGeneratorAgent(config, phase, runDir),
-    new UnitRunnerAgent(config, phase, runDir),
-    new BrowserCrawlerAgent(config, phase, runDir),
-    new ApiExerciserAgent(config, phase, runDir),
-    new SecurityScoutAgent(config, phase, runDir),
-    new A11yGuardianAgent(config, phase, runDir),
-    new HealerAgent(config, phase, runDir),
-    new ReporterAgent(config, phase, runDir),
-    new FixerAgent(config, phase, runDir),
-    new UxInspectorAgent(config, phase, runDir),
-    new PerformanceProfilerAgent(config, phase, runDir),
-    new DataIntegrityAuditorAgent(config, phase, runDir),
-    new RegressionSentinelAgent(config, phase, runDir),
-    new ChaosAgent(config, phase, runDir),
-    new DocumentationValidatorAgent(config, phase, runDir),
-    new IntegrationWatchdogAgent(config, phase, runDir),
-    new TenantIsolationAuditorAgent(config, phase, runDir),
-    new WorkflowCompletionTesterAgent(config, phase, runDir),
-    new StateSessionTesterAgent(config, phase, runDir),
-    new EmailNotificationVerifierAgent(config, phase, runDir),
-    new MigrationTesterAgent(config, phase, runDir),
-    new SignupOnboardingTesterAgent(config, phase, runDir),
-    new CrudFlowTesterAgent(config, phase, runDir),
-    new FormValidatorAgent(config, phase, runDir),
-    new SearchFilterTesterAgent(config, phase, runDir),
-    new NavigationRoutingTesterAgent(config, phase, runDir),
-    new ResponsiveInteractionTesterAgent(config, phase, runDir),
-    new MultiUserScenarioTesterAgent(config, phase, runDir),
-    new LoadTesterAgent(config, phase, runDir),
-    new MemoryLeakDetectorAgent(config, phase, runDir),
-    new BundleAnalyzerAgent(config, phase, runDir),
-    new XssScannerAgent(config, phase, runDir),
-    new CsrfTesterAgent(config, phase, runDir),
-    new AuthFuzzerAgent(config, phase, runDir),
-    new DependencyScannerAgent(config, phase, runDir),
-    new SecretsScannerAgent(config, phase, runDir),
-    new ApiContractTesterAgent(config, phase, runDir),
-    new RateLimitTesterAgent(config, phase, runDir),
-    new ApiPaginationTesterAgent(config, phase, runDir),
-    new GraphqlTesterAgent(config, phase, runDir),
-    new DataConsistencyCheckerAgent(config, phase, runDir),
-    new BackupRecoveryTesterAgent(config, phase, runDir),
-    new DataPrivacyScannerAgent(config, phase, runDir),
-    new SeoAuditorAgent(config, phase, runDir),
-    new SocialPreviewTesterAgent(config, phase, runDir),
-    new LighthouseAuditorAgent(config, phase, runDir),
-    new I18nTesterAgent(config, phase, runDir),
-    new TimezoneTesterAgent(config, phase, runDir),
-    new ErrorRecoveryTesterAgent(config, phase, runDir),
-    new OfflineModeTesterAgent(config, phase, runDir),
-    new GracefulDegradationTesterAgent(config, phase, runDir),
-    new WebSocketTesterAgent(config, phase, runDir),
-    new RealtimeSyncTesterAgent(config, phase, runDir),
-    new FileUploadTesterAgent(config, phase, runDir),
-    new ExportTesterAgent(config, phase, runDir),
-    new PaymentFlowTesterAgent(config, phase, runDir),
-    new SslTlsAuditorAgent(config, phase, runDir),
-    new DnsCdnTesterAgent(config, phase, runDir),
-    new DockerHealthCheckerAgent(config, phase, runDir),
-    new EnvConfigValidatorAgent(config, phase, runDir),
-    new LogQualityAuditorAgent(config, phase, runDir),
-    new AnalyticsTrackerTesterAgent(config, phase, runDir),
-    new GdprComplianceTesterAgent(config, phase, runDir),
-    new Soc2ControlValidatorAgent(config, phase, runDir),
-    new WcagAaaTesterAgent(config, phase, runDir),
-    new DeadCodeDetectorAgent(config, phase, runDir),
-    new TypeSafetyAuditorAgent(config, phase, runDir),
-    new ComplexityAnalyzerAgent(config, phase, runDir),
+    new AnalystAgent(config, phase, runDir, totalRecall),
+    new SeedArchitectAgent(config, phase, runDir, totalRecall),
+    new TestGeneratorAgent(config, phase, runDir, totalRecall),
+    new UnitRunnerAgent(config, phase, runDir, totalRecall),
+    new BrowserCrawlerAgent(config, phase, runDir, totalRecall),
+    new ApiExerciserAgent(config, phase, runDir, totalRecall),
+    new SecurityScoutAgent(config, phase, runDir, totalRecall),
+    new A11yGuardianAgent(config, phase, runDir, totalRecall),
+    new HealerAgent(config, phase, runDir, totalRecall),
+    new ReporterAgent(config, phase, runDir, totalRecall),
+    new FixerAgent(config, phase, runDir, totalRecall),
+    new UxInspectorAgent(config, phase, runDir, totalRecall),
+    new PerformanceProfilerAgent(config, phase, runDir, totalRecall),
+    new DataIntegrityAuditorAgent(config, phase, runDir, totalRecall),
+    new RegressionSentinelAgent(config, phase, runDir, totalRecall),
+    new ChaosAgent(config, phase, runDir, totalRecall),
+    new DocumentationValidatorAgent(config, phase, runDir, totalRecall),
+    new IntegrationWatchdogAgent(config, phase, runDir, totalRecall),
+    new TenantIsolationAuditorAgent(config, phase, runDir, totalRecall),
+    new WorkflowCompletionTesterAgent(config, phase, runDir, totalRecall),
+    new StateSessionTesterAgent(config, phase, runDir, totalRecall),
+    new EmailNotificationVerifierAgent(config, phase, runDir, totalRecall),
+    new MigrationTesterAgent(config, phase, runDir, totalRecall),
+    new SignupOnboardingTesterAgent(config, phase, runDir, totalRecall),
+    new CrudFlowTesterAgent(config, phase, runDir, totalRecall),
+    new FormValidatorAgent(config, phase, runDir, totalRecall),
+    new SearchFilterTesterAgent(config, phase, runDir, totalRecall),
+    new NavigationRoutingTesterAgent(config, phase, runDir, totalRecall),
+    new ResponsiveInteractionTesterAgent(config, phase, runDir, totalRecall),
+    new MultiUserScenarioTesterAgent(config, phase, runDir, totalRecall),
+    new LoadTesterAgent(config, phase, runDir, totalRecall),
+    new MemoryLeakDetectorAgent(config, phase, runDir, totalRecall),
+    new BundleAnalyzerAgent(config, phase, runDir, totalRecall),
+    new XssScannerAgent(config, phase, runDir, totalRecall),
+    new CsrfTesterAgent(config, phase, runDir, totalRecall),
+    new AuthFuzzerAgent(config, phase, runDir, totalRecall),
+    new DependencyScannerAgent(config, phase, runDir, totalRecall),
+    new SecretsScannerAgent(config, phase, runDir, totalRecall),
+    new ApiContractTesterAgent(config, phase, runDir, totalRecall),
+    new RateLimitTesterAgent(config, phase, runDir, totalRecall),
+    new ApiPaginationTesterAgent(config, phase, runDir, totalRecall),
+    new GraphqlTesterAgent(config, phase, runDir, totalRecall),
+    new DataConsistencyCheckerAgent(config, phase, runDir, totalRecall),
+    new BackupRecoveryTesterAgent(config, phase, runDir, totalRecall),
+    new DataPrivacyScannerAgent(config, phase, runDir, totalRecall),
+    new SeoAuditorAgent(config, phase, runDir, totalRecall),
+    new SocialPreviewTesterAgent(config, phase, runDir, totalRecall),
+    new LighthouseAuditorAgent(config, phase, runDir, totalRecall),
+    new I18nTesterAgent(config, phase, runDir, totalRecall),
+    new TimezoneTesterAgent(config, phase, runDir, totalRecall),
+    new ErrorRecoveryTesterAgent(config, phase, runDir, totalRecall),
+    new OfflineModeTesterAgent(config, phase, runDir, totalRecall),
+    new GracefulDegradationTesterAgent(config, phase, runDir, totalRecall),
+    new WebSocketTesterAgent(config, phase, runDir, totalRecall),
+    new RealtimeSyncTesterAgent(config, phase, runDir, totalRecall),
+    new FileUploadTesterAgent(config, phase, runDir, totalRecall),
+    new ExportTesterAgent(config, phase, runDir, totalRecall),
+    new PaymentFlowTesterAgent(config, phase, runDir, totalRecall),
+    new SslTlsAuditorAgent(config, phase, runDir, totalRecall),
+    new DnsCdnTesterAgent(config, phase, runDir, totalRecall),
+    new DockerHealthCheckerAgent(config, phase, runDir, totalRecall),
+    new EnvConfigValidatorAgent(config, phase, runDir, totalRecall),
+    new LogQualityAuditorAgent(config, phase, runDir, totalRecall),
+    new AnalyticsTrackerTesterAgent(config, phase, runDir, totalRecall),
+    new GdprComplianceTesterAgent(config, phase, runDir, totalRecall),
+    new Soc2ControlValidatorAgent(config, phase, runDir, totalRecall),
+    new WcagAaaTesterAgent(config, phase, runDir, totalRecall),
+    new DeadCodeDetectorAgent(config, phase, runDir, totalRecall),
+    new TypeSafetyAuditorAgent(config, phase, runDir, totalRecall),
+    new ComplexityAnalyzerAgent(config, phase, runDir, totalRecall),
   ];
 
   for (const agent of agents) {

package/core/cli.ts

@@ -6,6 +6,8 @@ import { registerAllAgents } from '../agents/registry';
 import { runInitWizard } from './init';
 import { determineExitCode, writeCISummary, detectCIEnvironment } from './ci-output';
 import type { ExitCode } from './ci-output';
+import { TotalRecallClient } from '../clients/total-recall';
+import { AgentMvpClient } from '../clients/agent-mvp';
 import * as path from 'path';
 
 const PHASES: readonly Phase[] = ['alpha', 'beta', 'uat', 'release'];
@@ -15,6 +17,9 @@ interface ParsedArgs {
   dryRun: boolean;
   agent: number | null;
   configPath: string;
+  recallQuery: string | null;
+  reputation: boolean;
+  coaching: string | null;
 }
 
 function printUsage(): void {
@@ -27,6 +32,9 @@ Options:
   --dry-run                              Show dispatch plan without executing
   --agent <id>                           Run a single agent by ID
   --config <path>                        Path to config file (default: testteam.config.ts)
+  --recall <query>                       Query Total Recall memory and print results
+  --reputation                           Query agent reputations and print a table
+  --coaching <agentId>                   Query coaching flags for an agent
   --help                                 Show this help message
 `.trim());
 }
@@ -37,6 +45,9 @@ export function parseArgs(argv: string[]): ParsedArgs {
   let dryRun = false;
   let agent: number | null = null;
   let configPath = 'testteam.config.ts';
+  let recallQuery: string | null = null;
+  let reputation = false;
+  let coaching: string | null = null;
 
   for (let i = 0; i < args.length; i++) {
     const arg = args[i];
@@ -67,6 +78,25 @@ export function parseArgs(argv: string[]): ParsedArgs {
           throw new Error('--config requires a path argument');
         }
         break;
+      case '--recall': {
+        const value = args[++i];
+        if (!value) {
+          throw new Error('--recall requires a query argument');
+        }
+        recallQuery = value;
+        break;
+      }
+      case '--reputation':
+        reputation = true;
+        break;
+      case '--coaching': {
+        const value = args[++i];
+        if (!value) {
+          throw new Error('--coaching requires an agent ID argument');
+        }
+        coaching = value;
+        break;
+      }
       case '--help':
         printUsage();
         process.exit(0);
@@ -76,7 +106,16 @@ export function parseArgs(argv: string[]): ParsedArgs {
     }
   }
 
-  return { phase, dryRun, agent, configPath };
+  return { phase, dryRun, agent, configPath, recallQuery, reputation, coaching };
+}
+
+async function checkMahAvailability(baseUrl = 'http://localhost:3300'): Promise<boolean> {
+  try {
+    const res = await fetch(`${baseUrl}/health`);
+    return res.ok;
+  } catch {
+    return false;
+  }
 }
 
 async function loadConfig(configPath: string): Promise<TestTeamConfig> {
@@ -112,7 +151,54 @@ export async function main(argv: string[] = process.argv): Promise<number> {
     return 1;
   }
 
-  const { phase, dryRun, configPath } = parsed;
+  const { phase, dryRun, configPath, recallQuery, reputation, coaching } = parsed;
+
+  // Handle --recall: query Total Recall and exit
+  if (recallQuery) {
+    const tr = new TotalRecallClient();
+    const available = await tr.isAvailable();
+    if (!available) {
+      console.error('Total Recall is not running at http://localhost:3777');
+      return 1;
+    }
+    const results = await tr.recallFindings(recallQuery);
+    console.log(JSON.stringify(results, null, 2));
+    return 0;
+  }
+
+  // Handle --coaching: query Agent-MVP coaching flags for a specific agent and exit
+  if (coaching) {
+    const mvp = new AgentMvpClient();
+    const mvpAvailable = await mvp.isAvailable();
+    if (!mvpAvailable) {
+      console.error('Agent-MVP is not running at http://localhost:3200');
+      return 1;
+    }
+    const flags = await mvp.getCoachingFlags(coaching);
+    if (flags.length > 0) {
+      console.log(JSON.stringify(flags, null, 2));
+    } else {
+      console.log(`No coaching flags for agent "${coaching}".`);
+    }
+    return 0;
+  }
+
+  // Handle --reputation: query Agent-MVP for all agent reputations and exit
+  if (reputation) {
+    const mvp = new AgentMvpClient();
+    const mvpAvailable = await mvp.isAvailable();
+    if (!mvpAvailable) {
+      console.error('Agent-MVP is not running at http://localhost:3200');
+      return 1;
+    }
+    const leaderboard = await mvp.getReputation('leaderboard');
+    if (leaderboard) {
+      console.log(JSON.stringify(leaderboard, null, 2));
+    } else {
+      console.log('No reputation data available.');
+    }
+    return 0;
+  }
 
   if (!dryRun && !phase) {
     console.error('--phase is required unless using --dry-run');
@@ -120,6 +206,10 @@ export async function main(argv: string[] = process.argv): Promise<number> {
     return 1;
   }
 
+  // Startup banner
+  console.log('TestTeam v3.0.0 — Learning Agents');
+  console.log('─'.repeat(40));
+
   // Load and validate config
   let rawConfig: TestTeamConfig;
   try {
@@ -131,6 +221,33 @@ export async function main(argv: string[] = process.argv): Promise<number> {
 
   const config = validateConfig(rawConfig);
 
+  // Initialize Total Recall (optional — continues normally if unavailable)
+  const totalRecall = new TotalRecallClient();
+  const trAvailable = await totalRecall.isAvailable();
+  if (trAvailable) {
+    await totalRecall.startSession();
+    console.log('[Total Recall] Connected — findings will be stored in memory');
+  } else {
+    console.log('[Total Recall] Not available at http://localhost:3777 — running without memory');
+  }
+
+  // Initialize Agent-MVP client (optional ��� continues normally if unavailable)
+  const agentMvp = new AgentMvpClient();
+  const mvpAvailable = await agentMvp.isAvailable();
+  if (mvpAvailable) {
+    console.log('[Agent-MVP] Connected — agent scores will be reported');
+  } else {
+    console.log('[Agent-MVP] Not available at http://localhost:3200 — running without scoring');
+  }
+
+  // Check MAH availability (optional — informational only)
+  const mahAvailable = await checkMahAvailability();
+  if (mahAvailable) {
+    console.log('[MAH] Connected — multi-agent hub active');
+  } else {
+    console.log('[MAH] Not available at http://localhost:3300 — running without hub');
+  }
+
   // Shared run directory so all phases write results to the same location.
   // This ensures the Reporter (Agent 10, release phase) can read results
   // from prior phases (alpha, beta, uat).
@@ -142,8 +259,8 @@ export async function main(argv: string[] = process.argv): Promise<number> {
       : [phase];
 
     for (const p of targetPhases) {
-      const phaseOrchestrator = new Orchestrator(config, sharedRunDir);
-      registerAllAgents(phaseOrchestrator, config, p, sharedRunDir);
+      const phaseOrchestrator = new Orchestrator(config, sharedRunDir, mvpAvailable ? agentMvp : undefined);
+      registerAllAgents(phaseOrchestrator, config, p, sharedRunDir, trAvailable ? totalRecall : undefined);
       const plan = phaseOrchestrator.dryRun(p);
       console.log(`\nPhase: ${p}`);
       for (const entry of plan) {
@@ -166,8 +283,8 @@ export async function main(argv: string[] = process.argv): Promise<number> {
   for (const p of targetPhases) {
     console.log(`\nRunning phase: ${p}`);
     try {
-      const phaseOrchestrator = new Orchestrator(config, sharedRunDir);
-      registerAllAgents(phaseOrchestrator, config, p, sharedRunDir);
+      const phaseOrchestrator = new Orchestrator(config, sharedRunDir, mvpAvailable ? agentMvp : undefined);
+      registerAllAgents(phaseOrchestrator, config, p, sharedRunDir, trAvailable ? totalRecall : undefined);
       const result = await phaseOrchestrator.runPhase(p);
 
       if (result.skippedAgents.length > 0) {
@@ -195,6 +312,11 @@ export async function main(argv: string[] = process.argv): Promise<number> {
     }
   }
 
+  // Close Total Recall session
+  if (trAvailable) {
+    await totalRecall.endSession().catch(() => {});
+  }
+
   // Determine structured exit code:
   // - 0: All gates passed, no advisory failures
   // - 1: Blocking agent failures (or runtime errors)

package/core/orchestrator.ts

@@ -11,6 +11,7 @@ import { getLicenseTier } from './license';
 import type { LicenseTier } from './license';
 import { isAgentAvailable, getAgentTier } from '../agents/registry';
 import { agentGatedMessage, phaseGatedMessage, getAgentDisplayName } from './messages';
+import type { AgentMvpClient, AgentScoreEvent } from '../clients/agent-mvp';
 import * as path from 'path';
 
 /** Agent IDs to dispatch per phase, in execution order. */
@@ -66,6 +67,7 @@ export class Orchestrator {
   private readonly agents: Map<number, BaseAgent> = new Map();
   private readonly healthChecks: HealthCheck[] = [];
   private runDir: string;
+  private readonly agentMvpClient: AgentMvpClient | null;
 
   /**
    * License tier resolver — injectable for testing.
@@ -73,9 +75,10 @@ export class Orchestrator {
    */
   _getLicenseTier: () => LicenseTier = () => getLicenseTier();
 
-  constructor(config: ValidatedConfig, runDir?: string) {
+  constructor(config: ValidatedConfig, runDir?: string, agentMvpClient?: AgentMvpClient) {
     this.config = config;
     this.runDir = runDir ?? path.join(process.cwd(), '.run', `run-${Date.now()}`);
+    this.agentMvpClient = agentMvpClient ?? null;
   }
 
   /** Register an agent instance for a given agent ID. */
@@ -282,6 +285,9 @@ export class Orchestrator {
       scheduledForPhase,
     );
 
+    // Step 7: Score agents in Agent-MVP (if connected)
+    await this.scoreAgentsInMvp(phase, agentResults);
+
     return { gate, agentResults, skippedAgents, costTracker };
   }
 
@@ -349,6 +355,71 @@ export class Orchestrator {
     }
   }
 
+  /**
+   * After a phase completes, send scoring events to Agent-MVP for each agent
+   * that produced results. Passed agents receive +1.0, failed agents receive -0.5.
+   */
+  private async scoreAgentsInMvp(
+    phase: Phase,
+    agentResults: Record<number, AgentResult>,
+  ): Promise<void> {
+    if (!this.agentMvpClient) return;
+
+    const events: AgentScoreEvent[] = [];
+
+    for (const [agentId, result] of Object.entries(agentResults)) {
+      if (result.status === 'skipped') continue;
+
+      const isPassed = result.status === 'passed';
+      const findingsCount = result.findings.length;
+      const worstSeverity = this.pickWorstSeverity(result.findings);
+
+      events.push({
+        agent_id: `agent-${agentId}`,
+        task_type: result.agentName,
+        outcome: isPassed ? 'success' : 'failure',
+        score_delta: isPassed ? 1.0 : -0.5,
+        phase,
+        findings_count: findingsCount > 0 ? findingsCount : undefined,
+        finding_severity: worstSeverity ?? undefined,
+      });
+    }
+
+    if (events.length > 0) {
+      try {
+        await this.agentMvpClient.scoreAgents(events);
+      } catch {
+        // Graceful -- Agent-MVP scoring is non-critical
+      }
+    }
+  }
+
+  /** Determine the worst (most severe) finding severity from a list. */
+  private pickWorstSeverity(findings: Finding[]): string | null {
+    if (findings.length === 0) return null;
+
+    const rank: Record<string, number> = {
+      critical: 5,
+      high: 4,
+      medium: 3,
+      low: 2,
+      info: 1,
+    };
+
+    let worst: string | null = null;
+    let worstRank = 0;
+
+    for (const finding of findings) {
+      const currentRank = rank[finding.severity] ?? 0;
+      if (currentRank > worstRank) {
+        worstRank = currentRank;
+        worst = finding.severity;
+      }
+    }
+
+    return worst;
+  }
+
   /** Get the run directory path. */
   getRunDir(): string {
     return this.runDir;

package/dist/agents/15-regression-sentinel.d.ts

@@ -1,12 +1,13 @@
 import type { Finding, Phase } from '../core/types';
 import type { ValidatedConfig } from '../core/config';
+import type { TotalRecallClient } from '../clients/total-recall';
 import { BaseAgent } from './base-agent';
 export declare class RegressionSentinelAgent extends BaseAgent {
     readonly agentId = 15;
     readonly agentName = "Regression Sentinel";
     /** Absolute path to the baselines directory (sibling of runDir). */
     private readonly baselinesDir;
-    constructor(config: ValidatedConfig, phase: Phase, runDir: string);
+    constructor(config: ValidatedConfig, phase: Phase, runDir: string, totalRecall?: TotalRecallClient);
     protected preFlight(): Promise<void>;
     protected execute(): Promise<Finding[]>;
     /**

package/dist/agents/15-regression-sentinel.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"15-regression-sentinel.d.ts","sourceRoot":"","sources":["../../agents/15-regression-sentinel.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,eAAe,CAAC;AACpD,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,gBAAgB,CAAC;AACtD,OAAO,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAkBzC,qBAAa,uBAAwB,SAAQ,SAAS;IACpD,QAAQ,CAAC,OAAO,MAAM;IACtB,QAAQ,CAAC,SAAS,yBAAyB;IAE3C,oEAAoE;IACpE,OAAO,CAAC,QAAQ,CAAC,YAAY,CAAS;gBAE1B,MAAM,EAAE,eAAe,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM;cAOjD,SAAS,IAAI,OAAO,CAAC,IAAI,CAAC;cAa1B,OAAO,IAAI,OAAO,CAAC,OAAO,EAAE,CAAC;IAuI7C;;;;;;;OAOG;YACW,UAAU;IAiBxB,iFAAiF;IACjF,OAAO,CAAC,WAAW;YAOL,iBAAiB;CAyFhC"}
+{"version":3,"file":"15-regression-sentinel.d.ts","sourceRoot":"","sources":["../../agents/15-regression-sentinel.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,eAAe,CAAC;AACpD,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,gBAAgB,CAAC;AACtD,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AACjE,OAAO,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAkBzC,qBAAa,uBAAwB,SAAQ,SAAS;IACpD,QAAQ,CAAC,OAAO,MAAM;IACtB,QAAQ,CAAC,SAAS,yBAAyB;IAE3C,oEAAoE;IACpE,OAAO,CAAC,QAAQ,CAAC,YAAY,CAAS;gBAE1B,MAAM,EAAE,eAAe,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,WAAW,CAAC,EAAE,iBAAiB;cAOlF,SAAS,IAAI,OAAO,CAAC,IAAI,CAAC;cAa1B,OAAO,IAAI,OAAO,CAAC,OAAO,EAAE,CAAC;IAuI7C;;;;;;;OAOG;YACW,UAAU;IAiBxB,iFAAiF;IACjF,OAAO,CAAC,WAAW;YAOL,iBAAiB;CAyFhC"}

package/dist/agents/15-regression-sentinel.js

@@ -20,8 +20,8 @@ export class RegressionSentinelAgent extends BaseAgent {
     agentName = 'Regression Sentinel';
     /** Absolute path to the baselines directory (sibling of runDir). */
     baselinesDir;
-    constructor(config, phase, runDir) {
-        super(config, phase, runDir);
+    constructor(config, phase, runDir, totalRecall) {
+        super(config, phase, runDir, totalRecall);
         // Place baselines as a sibling directory to runDir
         const parentDir = join(runDir, '..');
         this.baselinesDir = join(parentDir, BASELINES_DIR_NAME);