thumbgate 1.30.0 → 1.34.0

package/public/pricing.html

@@ -4,7 +4,7 @@
   <meta charset="utf-8">
   <meta name="viewport" content="width=device-width, initial-scale=1">
   <title>Pricing — Enforcement that pays for itself | ThumbGate</title>
-  <meta name="description" content="ThumbGate vs DIY prompts, platform hire, or GRC suite. Free evaluate, Pro $19/mo, $499 Diagnostic gate. Countable units that convert.">
+  <meta name="description" content="Choose ThumbGate by the value and scope of the agent risk: free evaluation, Pro for one operator, a $499 managed diagnostic for one workflow, or custom enterprise intake.">
   <link rel="canonical" href="__APP_ORIGIN__/pricing">
   <link rel="alternate" type="text/markdown" title="ThumbGate LLM context" href="__APP_ORIGIN__/llm-context.md">
   <meta property="og:title" content="ThumbGate Pricing — $499 Diagnostic · Pro $19/mo">
@@ -57,6 +57,19 @@
     .fit-card { border-top:3px solid var(--green); background:var(--card); padding:22px; }
     .fit-card h3 { margin:0 0 8px; }
     .fit-card p { margin:0; color:var(--muted); }
+    .value-equation { display:grid; grid-template-columns:1fr auto 1fr auto 1fr auto 1.3fr; gap:12px; align-items:stretch; margin-top:26px; }
+    .value-factor { display:grid; align-content:center; min-height:112px; border:1px solid var(--line); border-radius:14px; background:var(--card); padding:18px; }
+    .value-factor strong { font-size:1.06rem; }
+    .value-factor span { color:var(--muted); font-size:.83rem; }
+    .value-operator { display:grid; place-items:center; color:var(--green); font-size:1.7rem; font-weight:900; }
+    .value-result { border-color:var(--green); background:#e9f3ee; }
+    .segment-grid { display:grid; grid-template-columns:repeat(3,1fr); gap:14px; margin-top:28px; }
+    .segment-card { display:flex; flex-direction:column; border:1px solid var(--line); border-radius:14px; background:var(--card); padding:22px; }
+    .segment-card.recommended { border:2px solid var(--green); }
+    .segment-card h3 { margin:8px 0 6px; }
+    .segment-card p { margin:0 0 14px; color:var(--muted); font-size:.92rem; }
+    .segment-card .fence { min-height:58px; margin-top:auto; color:var(--ink); font-size:.82rem; }
+    .segment-card .button { margin-top:14px; }
     .faq { border-top:1px solid var(--line); }
     .faq-item { border-bottom:1px solid var(--line); }
     .faq-q { width:100%; border:0; background:transparent; padding:18px 0; color:var(--ink); font:inherit; font-weight:800; text-align:left; cursor:pointer; }
@@ -65,22 +78,52 @@
     footer { border-top:1px solid var(--line); padding:28px 0; color:var(--muted); font-size:.85rem; }
     .footer-inner { display:flex; justify-content:space-between; gap:24px; }
     .footer-links { display:flex; gap:18px; }
-    @media (max-width:820px) { .hero,.steps,.fit { grid-template-columns:1fr; } .hero { gap:34px; } .nav-links a:not(.nav-buy) { display:none; } }
+    @media (max-width:820px) { .hero,.steps,.fit,.segment-grid { grid-template-columns:1fr; } .value-equation { grid-template-columns:1fr; } .value-operator { min-height:24px; } .hero { gap:34px; } .nav-links a:not(.nav-buy) { display:none; } }
   </style>
   <script type="application/ld+json">
   {
     "@context": "https://schema.org",
-    "@type": "Service",
-    "name": "ThumbGate Enterprise Workflow Gate",
-    "description": "A bounded implementation for one supported local AI-agent workflow: review, configured pre-action gate, regression test, and rollout and rollback proof.",
-    "provider": { "@type": "Organization", "name": "ThumbGate", "url": "__APP_ORIGIN__/" },
-    "offers": {
-      "@type": "Offer",
-      "price": "499",
-      "priceCurrency": "USD",
-      "url": "__APP_ORIGIN__/pricing#buy",
-      "availability": "https://schema.org/InStock"
-    }
+    "@graph": [
+      {
+        "@type": "Service",
+        "name": "ThumbGate Managed Workflow Diagnostic",
+        "description": "A bounded implementation for one supported local AI-agent workflow: review, configured pre-action gate, regression test, and rollout and rollback proof.",
+        "provider": { "@type": "Organization", "name": "ThumbGate", "url": "__APP_ORIGIN__/" },
+        "audience": { "@type": "BusinessAudience", "audienceType": "Engineering and platform teams with a repeated AI-agent failure" },
+        "offers": {
+          "@type": "Offer",
+          "name": "Managed workflow diagnostic",
+          "price": "499",
+          "priceCurrency": "USD",
+          "url": "__APP_ORIGIN__/pricing#buy",
+          "availability": "https://schema.org/InStock"
+        }
+      },
+      {
+        "@type": "SoftwareApplication",
+        "name": "ThumbGate Pro",
+        "applicationCategory": "DeveloperApplication",
+        "operatingSystem": "macOS, Linux, Windows",
+        "offers": [
+          {
+            "@type": "Offer",
+            "name": "ThumbGate Pro monthly",
+            "price": "19",
+            "priceCurrency": "USD",
+            "url": "__APP_ORIGIN__/checkout/pro?plan_id=pro",
+            "availability": "https://schema.org/InStock"
+          },
+          {
+            "@type": "Offer",
+            "name": "ThumbGate Pro annual",
+            "price": "149",
+            "priceCurrency": "USD",
+            "url": "__APP_ORIGIN__/checkout/pro?plan_id=pro",
+            "availability": "https://schema.org/InStock"
+          }
+        ]
+      }
+    ]
   }
   </script>
   <script type="application/ld+json">
@@ -109,8 +152,8 @@
       <div class="nav-links">
         <a href="/guide">Technical setup</a>
         <a href="https://github.com/IgorGanapolsky/ThumbGate/blob/main/docs/VERIFICATION_EVIDENCE.md" target="_blank" rel="noopener">Proof</a>
-        <a class="nav-enterprise" href="#enterprise-gate" data-offer-link data-cta-id="pricing_nav_buy">Get Started · $499</a>
-        <a class="nav-pro" href="/checkout/pro?utm_source=pricing&amp;utm_medium=nav&amp;utm_campaign=rally_style_gtm&amp;cta_id=pricing_nav_pro&amp;plan_id=pro" data-offer-link data-cta-id="pricing_nav_pro">Pro · $19/mo</a>
+        <a class="nav-enterprise" href="#enterprise-gate" data-offer-link data-cta-id="pricing_nav_buy" data-plan-id="sprint_diagnostic" data-value="499" data-segment="workflow_team">Get Started · $499</a>
+        <a class="nav-pro" href="/checkout/pro?utm_source=pricing&amp;utm_medium=nav&amp;utm_campaign=value_packaging_v1&amp;cta_id=pricing_nav_pro&amp;plan_id=pro" data-offer-link data-cta-id="pricing_nav_pro" data-plan-id="pro" data-value="19" data-segment="solo_operator">Pro · $19/mo</a>
       </div>
     </div>
   </nav>
@@ -139,29 +182,74 @@
             <input type="hidden" name="plan_id" value="sprint_diagnostic">
             <input type="hidden" name="utm_source" value="pricing">
             <input type="hidden" name="utm_medium" value="direct_checkout">
-            <input type="hidden" name="utm_campaign" value="rally_style_gtm">
+            <input type="hidden" name="utm_campaign" value="value_packaging_v1">
+            <input id="pricing-buyer-segment" type="hidden" name="campaign_variant" value="workflow_team">
             <input type="hidden" name="cta_id" value="pricing_managed_gate_buy">
             <input type="hidden" name="cta_placement" value="pricing">
             <input type="hidden" name="landing_path" value="/pricing">
             <button class="button" type="submit">Get Started — $499 Diagnostic</button>
           </form>
           <p class="fine">Secure payment. We reply with scheduling and the workflow checklist.</p>
-          <div class="boundary">Detected secret exfiltration and gate-process bypass attempts deny by default. Other destructive actions warn by default and deny when strict enforcement is enabled.</div>
+          <div class="boundary">Detected secret exfiltration denies by default: literal secrets, secret-file paths in uploads/pipes/redirections/command-substitution, secret env vars to network tools, scp/rsync/cloud CLI uploads of credential files, and common interpreter one-liners. Not a full network sandbox. Other destructive actions warn by default and deny when strict enforcement is enabled.</div>
         </aside>
         <aside class="pro-card" id="pro" aria-label="ThumbGate Pro">
           <div class="eyebrow">Self-serve</div>
           <div class="price">$19<span style="font-size:1rem;font-weight:700">/mo</span></div>
           <p class="price-note">Or $149/yr. Individual operators. Cancel anytime.</p>
-          <a class="button" href="/checkout/pro?utm_source=pricing&amp;utm_medium=card&amp;utm_campaign=rally_style_gtm&amp;cta_id=pricing_pro_buy&amp;plan_id=pro" data-offer-link data-cta-id="pricing_pro_buy" style="width:100%;box-sizing:border-box">Start Pro — $19/mo</a>
+          <a class="button" href="/checkout/pro?utm_source=pricing&amp;utm_medium=card&amp;utm_campaign=value_packaging_v1&amp;cta_id=pricing_pro_buy&amp;plan_id=pro" data-offer-link data-cta-id="pricing_pro_buy" data-plan-id="pro" data-value="19" data-segment="solo_operator" style="width:100%;box-sizing:border-box">Start Pro</a>
           <p class="fine">Hosted checkout. Free local evaluate remains free after you subscribe.</p>
         </aside>
       </div>
     </div>
 
+    <section id="value">
+      <div class="eyebrow">Price from your exposure</div>
+      <h2>Value is the recovery work the next repeat does not create.</h2>
+      <p class="section-lede">Use your own incident history. ThumbGate does not claim a universal savings number.</p>
+      <div class="value-equation" role="img" aria-label="Repeat incidents multiplied by recovery hours and loaded hourly cost, plus downstream impact, equals avoidable recovery exposure">
+        <div class="value-factor"><strong>Repeat incidents</strong><span>per quarter</span></div>
+        <div class="value-operator">×</div>
+        <div class="value-factor"><strong>Recovery hours</strong><span>diagnosis, repair, review</span></div>
+        <div class="value-operator">×</div>
+        <div class="value-factor"><strong>Loaded hourly cost</strong><span>the people pulled into recovery</span></div>
+        <div class="value-operator">+</div>
+        <div class="value-factor value-result"><strong>Downstream impact</strong><span>delay, incident response, audit evidence</span></div>
+      </div>
+    </section>
+
+    <section id="segments">
+      <div class="eyebrow">Choose by buyer value</div>
+      <h2>One product. Three buying decisions.</h2>
+      <p class="section-lede">The fence is scope and service level—not an arbitrary discount.</p>
+      <div class="segment-grid">
+        <article class="segment-card">
+          <div class="eyebrow">Solo operator</div>
+          <h3>Evaluate free, then Pro</h3>
+          <p>Pro removes rule limits and adds personal recall, review, dashboard, and export workflows.</p>
+          <div class="fence"><strong>Fence:</strong> one operator, local-first, self-serve support. $19 monthly or $149 annual.</div>
+          <a class="button" href="/checkout/pro?utm_source=pricing&amp;utm_medium=segment_card&amp;utm_campaign=value_packaging_v1&amp;cta_id=pricing_segment_pro&amp;plan_id=pro" data-offer-link data-cta-id="pricing_segment_pro" data-plan-id="pro" data-value="19" data-segment="solo_operator">Choose Pro</a>
+        </article>
+        <article class="segment-card recommended">
+          <div class="eyebrow">Engineering or platform team</div>
+          <h3>Managed diagnostic</h3>
+          <p>Turn one expensive, repeated agent failure into a configured gate and a regression proof.</p>
+          <div class="fence"><strong>Fence:</strong> one supported local workflow, fixed scope, two-business-day delivery. $499 once.</div>
+          <a class="button" href="#enterprise-gate" data-offer-link data-cta-id="pricing_segment_diagnostic" data-plan-id="sprint_diagnostic" data-value="499" data-segment="workflow_team">Harden one workflow</a>
+        </article>
+        <article class="segment-card">
+          <div class="eyebrow">Regulated or multi-workflow team</div>
+          <h3>Enterprise service intake</h3>
+          <p>Scope approval boundaries, rollback evidence, and accountable rollout support after one workflow proves fit.</p>
+          <div class="fence"><strong>Fence:</strong> custom scope after intake. Hosted team features, SSO, and SIEM are not general availability.</div>
+          <a class="button" href="#enterprise-gate" data-offer-link data-cta-id="pricing_segment_enterprise" data-plan-id="enterprise_service" data-value="0" data-segment="regulated_team">Start with one diagnostic</a>
+        </article>
+      </div>
+    </section>
+
     <section>
       <div class="eyebrow">How ThumbGate stacks up</div>
       <h2>Make the alternative look expensive.</h2>
-      <p class="section-lede">Same move as high-converting SaaS pricing: DIY prompts cost nothing and stop nothing; platform hire and GRC suites cost months.</p>
+      <p class="section-lede">DIY prompts look cheap but still consume model and review time. Platform hiring and GRC suites require larger commitments.</p>
       <div style="overflow-x:auto;margin-top:22px;border:1px solid var(--line);border-radius:14px;background:var(--card);">
         <table style="width:100%;border-collapse:collapse;font-size:.92rem;min-width:620px;">
           <thead>
@@ -176,24 +264,24 @@
           <tbody>
             <tr>
               <td style="padding:12px 14px;border-bottom:1px solid var(--line);color:var(--muted);">Monthly cost</td>
-              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">$0</td>
-              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">$15k+ loaded</td>
-              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">$$$$</td>
+              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Model + review time</td>
+              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Loaded engineering cost</td>
+              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Contract + implementation</td>
               <td style="padding:12px 14px;border-bottom:1px solid var(--line);font-weight:800;color:var(--green);">$0–$19 · $499 once</td>
             </tr>
             <tr>
               <td style="padding:12px 14px;border-bottom:1px solid var(--line);color:var(--muted);">Time to first gate</td>
-              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Never reliable</td>
+              <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Manual upkeep</td>
               <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Weeks</td>
               <td style="padding:12px 14px;border-bottom:1px solid var(--line);">Months</td>
               <td style="padding:12px 14px;border-bottom:1px solid var(--line);font-weight:800;">Minutes after install / 2 business days managed</td>
             </tr>
             <tr>
               <td style="padding:12px 14px;color:var(--muted);">When agent “looks competent”</td>
-              <td style="padding:12px 14px;">Ignores the prompt</td>
+              <td style="padding:12px 14px;">Prompt may be skipped</td>
               <td style="padding:12px 14px;">Custom glue</td>
               <td style="padding:12px 14px;">Gateway only</td>
-              <td style="padding:12px 14px;font-weight:800;">Hard block at tool call</td>
+              <td style="padding:12px 14px;font-weight:800;">Configured gate at tool call</td>
             </tr>
           </tbody>
         </table>
@@ -254,13 +342,31 @@
       } catch (_) {}
     }
 
-    function offerProps(ctaId, placement) {
-      return { ctaId: ctaId, ctaPlacement: placement, planId: 'sprint_diagnostic', value: 499, currency: 'USD' };
+    function offerProps(ctaId, placement, planId, value, segment) {
+      return {
+        ctaId: ctaId,
+        ctaPlacement: placement,
+        planId: planId,
+        value: Number(value || 0),
+        currency: 'USD',
+        segment: segment,
+        experimentId: 'value_packaging_v1'
+      };
     }
 
     document.querySelectorAll('[data-offer-link]').forEach(function (link) {
       link.addEventListener('click', function () {
-        const props = offerProps(link.dataset.ctaId || 'pricing_anchor', 'pricing_anchor');
+        if (link.getAttribute('href') === '#enterprise-gate') {
+          const buyerSegment = document.querySelector('#pricing-buyer-segment');
+          if (buyerSegment) buyerSegment.value = link.dataset.segment || 'workflow_team';
+        }
+        const props = offerProps(
+          link.dataset.ctaId || 'pricing_anchor',
+          'pricing_anchor',
+          link.dataset.planId || 'unknown',
+          link.dataset.value || '0',
+          link.dataset.segment || 'unknown'
+        );
         window.plausible('pricing_cta_click', { props: props });
         sendFirstPartyTelemetry('cta_click', props);
       });
@@ -269,7 +375,14 @@
     const checkoutForm = document.querySelector('[data-primary-checkout]');
     if (checkoutForm) {
       checkoutForm.addEventListener('submit', function () {
-        const props = offerProps('pricing_managed_gate_buy', 'pricing');
+        const buyerSegment = document.querySelector('#pricing-buyer-segment');
+        const props = offerProps(
+          'pricing_managed_gate_buy',
+          'pricing',
+          'sprint_diagnostic',
+          499,
+          buyerSegment ? buyerSegment.value : 'workflow_team'
+        );
         window.plausible('checkout_start', { props: props });
         sendFirstPartyTelemetry('checkout_start', props);
       });
@@ -283,7 +396,7 @@
       });
     });
 
-    sendFirstPartyTelemetry('pricing_page_view', { landingPath: '/pricing' });
+    sendFirstPartyTelemetry('pricing_page_view', { landingPath: '/pricing', experimentId: 'value_packaging_v1' });
   </script>
 </body>
 </html>

package/scripts/a-plus-evidence-scorecard.js

@@ -0,0 +1,303 @@
+#!/usr/bin/env node
+'use strict';
+
+/**
+ * Fail-closed A+ readiness scorecard.
+ *
+ * A passing unit test or a checked-in module is repository evidence, not live
+ * production or commercial proof. This scorecard keeps those surfaces separate
+ * and awards A+/10 only when every check in every area is verified.
+ */
+
+const fs = require('node:fs');
+const path = require('node:path');
+
+const ROOT = path.join(__dirname, '..');
+const SCORECARD_VERSION = '2026-08-01.1';
+
+function read(root, relativePath) {
+  try {
+    return fs.readFileSync(path.join(root, relativePath), 'utf8');
+  } catch {
+    return '';
+  }
+}
+
+function hasAll(text, needles) {
+  return needles.every((needle) => text.includes(needle));
+}
+
+function safeEval(fn) {
+  try {
+    return fn() === true;
+  } catch {
+    return false;
+  }
+}
+
+function collectRepositoryEvidence(root = ROOT) {
+  const landing = read(root, 'public/index.html');
+  const feedback = read(root, 'scripts/feedback-loop.js');
+  const promotion = read(root, 'scripts/auto-promote-gates.js');
+  const buyerPaths = read(root, 'scripts/buyer-paths.js');
+  const hook = read(root, 'scripts/hook-pre-tool-use.js');
+  const gatesEngine = read(root, 'scripts/gates-engine.js');
+  const gateEvasionMatrix = read(root, 'tests/gate-evasion-matrix.test.js');
+  const retrieval = read(root, 'scripts/lesson-retrieval.js');
+  const crossEncoder = read(root, 'scripts/cross-encoder-reranker.js');
+  const productionArchitecture = read(root, 'docs/RAG_PRODUCTION_ARCHITECTURE.md');
+  const packageJson = read(root, 'package.json');
+
+  let qualitySuitePassed = false;
+  let rerankGoldenPassed = false;
+  if (root === ROOT) {
+    qualitySuitePassed = safeEval(() => require('./eval-quality-suite').runSuite().report.passed);
+    rerankGoldenPassed = safeEval(() => require('./rerank-quality-eval').evaluate().pass);
+  }
+
+  return {
+    landingVisualLoop: hasAll(landing, [
+      'hero-thumbs',
+      'before-after.svg',
+      'self-improving-thumbs-loop.svg',
+      'Is it really self-improving?',
+    ]),
+    landingBuyerRoutes: hasAll(landing, ['/checkout/pro', '/go/diagnostic-pay'])
+      && hasAll(buyerPaths, ['/go/pro', '/go/sprint', '/diagnostic']),
+    feedbackRewardReachable: feedback.includes('scoreFeedbackReward('),
+    feedbackPromotionReachable: promotion.includes('promote') && hook.includes('retrieveWithRerankingSync'),
+    preventionChangePromoted: feedback.includes('whatToChange')
+      && feedback.includes('promotion'),
+    architectureNamesHonest: hasAll(productionArchitecture, [
+      'LLM-as-a-judge output is diagnostic',
+      'A heuristic score is never',
+      'does not get to override a hard gate',
+    ]),
+    deterministicMultiQuery: retrieval.includes('buildQueryVariants'),
+    hydeExplicitAndBounded: hasAll(retrieval, ['hydeGenerator', 'hydeApplied', 'hydeProvider']),
+    rerankProductionWired: crossEncoder.includes("require('./rerank-pipeline')")
+      && crossEncoder.includes('rerankPipelineSync(query, candidates'),
+    rerankProvenance: hasAll(crossEncoder, [
+      'pairwiseHeuristicScore',
+      'crossEncoderScore',
+      'reranker',
+    ]),
+    rerankGoldenPassed,
+    qualitySuitePassed,
+    requestEnvelope: fs.existsSync(path.join(root, 'scripts/request-envelope.js')),
+    hardBudgets: fs.existsSync(path.join(root, 'scripts/tier-budget-guard.js')),
+    degradedRetrievalFlags: fs.existsSync(path.join(root, 'scripts/retrieval-quality-tier.js')),
+    structuredOutputValidation: fs.existsSync(path.join(root, 'scripts/rag-structured-output.js')),
+    tenantAclBeforeRetrieval: hasAll(productionArchitecture, [
+      'Filtering must happen before',
+      'Missing/mismatched tenant or principal',
+    ]),
+    commandPositionHardening: hasAll(gatesEngine, [
+      'LITERAL_COMMAND_SUBSTITUTION_HEADS',
+      'canonicalizeLiteralCommandSubstitutionHead',
+    ]) && gateEvasionMatrix.includes('literal command substitution'),
+    rawFrameworkDecisionDefended: hasAll(productionArchitecture, [
+      '## Framework decision',
+      'LangChain',
+      'LangGraph',
+      'LlamaIndex',
+      '## One complete RAG request',
+    ]),
+    scorecardInMainTest: packageJson.includes('test:a-plus-evidence'),
+  };
+}
+
+function check(id, label, passed, evidenceClass, remediation) {
+  return {
+    id,
+    label,
+    passed: passed === true,
+    evidenceClass,
+    remediation: passed === true ? null : remediation,
+  };
+}
+
+function shaMatches(live = {}) {
+  const candidate = String(live.candidateBuildSha || '').trim();
+  const deployed = String(live.deployedBuildSha || '').trim();
+  return candidate.length >= 7 && deployed.length >= 7 && candidate === deployed;
+}
+
+function evaluateReadiness({ repo = {}, live = {} } = {}) {
+  const production = live.production || {};
+  const retrieval = live.retrieval || {};
+  const security = live.security || {};
+  const commercial = live.commercial || {};
+
+  const areas = [
+    {
+      id: 'landing_conversion',
+      label: 'Landing page and conversion',
+      checks: [
+        check('visual_loop', 'Thumb visuals and simple learning diagrams ship', repo.landingVisualLoop, 'repository', 'Ship the visual thumbs-to-gate loop.'),
+        check('buyer_routes', 'First-party buyer routes are present', repo.landingBuyerRoutes, 'repository', 'Restore diagnostic, Pro, and sprint buyer routes.'),
+        check('live_landing', 'Candidate landing page is verified live', production.landingVerified === true && shaMatches(production), 'production', 'Verify the exact candidate SHA on the live landing page.'),
+      ],
+    },
+    {
+      id: 'self_improvement',
+      label: 'Self-improving control loop',
+      checks: [
+        check('reward_reachable', 'Feedback reward scoring is invoked by capture', repo.feedbackRewardReachable, 'repository', 'Wire reward scoring into the capture path.'),
+        check('promotion_reachable', 'Reviewed failures reach promotion and pre-action retrieval', repo.feedbackPromotionReachable, 'repository', 'Connect feedback promotion to the pre-action hook.'),
+        check('specific_change', 'Specific what-to-change guidance reaches prevention rules', repo.preventionChangePromoted, 'repository', 'Promote specific corrective instructions, not vague signals.'),
+        check('live_feedback', 'Fresh production feedback closes the loop', production.feedbackLoopVerified === true, 'production', 'Capture one real reviewed outcome and prove its next-action effect.'),
+      ],
+    },
+    {
+      id: 'architecture_honesty',
+      label: 'Judge, routing, and architecture honesty',
+      checks: [
+        check('honest_names', 'Judge, heuristic, neural, and enforcement stages are distinct', repo.architectureNamesHonest, 'repository', 'Document stage placement and prevent misleading model-level MoE claims.'),
+        check('route_trace', 'Live traces identify the provider and routed model', production.providerTraceVerified === true, 'production', 'Attach a secret-safe live route trace.'),
+      ],
+    },
+    {
+      id: 'query_transformation',
+      label: 'Query transformation, multi-query, and HyDE',
+      checks: [
+        check('multi_query', 'Bounded deterministic multi-query is implemented', repo.deterministicMultiQuery, 'repository', 'Implement bounded, inspectable query variants.'),
+        check('hyde_contract', 'HyDE is explicit, bounded, and provenance-bearing', repo.hydeExplicitAndBounded, 'repository', 'Add an explicit caller-supplied HyDE contract and fallback.'),
+        check('hyde_holdout', 'HyDE or multi-query improves a provider holdout', retrieval.queryTransformationHoldoutPassed === true, 'provider-holdout', 'Measure lift on a non-fixture provider holdout.'),
+      ],
+    },
+    {
+      id: 'reranking',
+      label: 'Reranking cascade',
+      checks: [
+        check('production_wiring', 'BM25F, local MaxSim, and pairwise fusion run in PreToolUse', repo.rerankProductionWired, 'repository', 'Wire the documented cascade into the production caller.'),
+        check('provenance', 'Heuristic and neural scores cannot masquerade as each other', repo.rerankProvenance, 'repository', 'Emit per-stage provenance and explicit fallbacks.'),
+        check('golden', 'Deterministic rerank golden floors pass', repo.rerankGoldenPassed, 'deterministic-eval', 'Fix rerank golden regressions.'),
+        check('neural_holdout', 'True neural pair/late-interaction holdout passes', retrieval.neuralRerankHoldoutPassed === true, 'provider-holdout', 'Run a pretrained pair scorer or token embedder on an external holdout.'),
+        check('llm_failures', 'LLM rerank failure modes pass live-provider tests', retrieval.llmRerankFailureModesPassed === true, 'provider-holdout', 'Test malformed, partial, injected, timed-out, and unavailable LLM reranks.'),
+      ],
+    },
+    {
+      id: 'evaluation',
+      label: 'Retrieval and answer evaluation',
+      checks: [
+        check('offline_suite', 'Recall, precision, MRR, nDCG, and answer proxy floors pass', repo.qualitySuitePassed, 'deterministic-eval', 'Fix the unified deterministic quality suite.'),
+        check('external_cases', 'External labeled holdout has at least 100 cases', Number(retrieval.externalHoldoutCases) >= 100, 'provider-holdout', 'Label and freeze at least 100 non-fixture cases.'),
+        check('judge_calibration', 'LLM judge is calibrated against human labels', retrieval.judgeCalibrationPassed === true, 'provider-holdout', 'Measure judge agreement and calibration against human-reviewed labels.'),
+      ],
+    },
+    {
+      id: 'production_controls',
+      label: 'Latency, cost, and observability',
+      checks: [
+        check('request_envelope', 'Request trace, token, cost, and retrieval envelope exists', repo.requestEnvelope, 'repository', 'Add a request envelope.'),
+        check('hard_budgets', 'Per-request and daily tier budgets fail closed', repo.hardBudgets, 'repository', 'Add hard cost and tier budgets.'),
+        check('degraded_flags', 'Stale or stub retrieval is labeled degraded', repo.degradedRetrievalFlags, 'repository', 'Expose retrieval quality tiers.'),
+        check('live_slo', 'Production p95 and cost SLOs pass under load', production.loadTestPassed === true && Number(production.p95LatencyMs) > 0, 'production', 'Run a production-like load test and attach p95/cost evidence.'),
+        check('cache_batch', 'Live cache and batching savings are measured', production.cacheAndBatchingMeasured === true, 'production', 'Measure cache hit rate and batching cost/latency lift.'),
+      ],
+    },
+    {
+      id: 'failure_security',
+      label: 'Failure modes, validation, ACL, and tenancy',
+      checks: [
+        check('structured', 'Structured answers and citations are validated', repo.structuredOutputValidation, 'repository', 'Validate output shape and citation relationships.'),
+        check('acl_order', 'Tenant/document ACL runs before retrieval and hydration', repo.tenantAclBeforeRetrieval, 'repository', 'Enforce authorization before ranking.'),
+        check('command_evasion', 'Literal command-substitution evasions are canonicalized and tested', repo.commandPositionHardening, 'repository', 'Ratchet deterministic command-position substitutions in the evasion matrix.'),
+        check('penetration_test', 'Tenant isolation has external penetration evidence', security.tenantPenTestPassed === true, 'security-review', 'Run a professional tenant-isolation penetration test.'),
+        check('incident_drill', 'Hallucination, stale-index, miss, and leak drills pass', security.failureDrillPassed === true, 'production', 'Run and retain production-like failure drills.'),
+      ],
+    },
+    {
+      id: 'framework_pipeline',
+      label: 'Framework decision and end-to-end RAG defense',
+      checks: [
+        check('decision', 'Raw versus LangChain/LangGraph/LlamaIndex tradeoffs are defended', repo.rawFrameworkDecisionDefended, 'repository', 'Document the complete pipeline and framework decision.'),
+        check('ratchet', 'The evidence scorecard runs in the main test chain', repo.scorecardInMainTest, 'repository', 'Wire this scorecard into the test chain.'),
+      ],
+    },
+    {
+      id: 'commercial_validation',
+      label: 'Value, willingness to pay, and captured money',
+      checks: [
+        check('buyer_conversations', 'At least 10 target-buyer value conversations are evidenced', Number(commercial.buyerConversations) >= 10, 'commercial', 'Complete and retain 10 target-buyer value conversations.'),
+        check('payment_asks', 'At least 3 exact-price payment asks are evidenced', Number(commercial.paymentAsks) >= 3, 'commercial', 'Make three exact-price payment asks to qualified buyers.'),
+        check('external_payment', 'At least one non-owner external payment is reconciled', Number(commercial.externalPayments) >= 1, 'provider', 'Capture and reconcile one real external payment.'),
+        check('provider_truth', 'Provider catalog and product attribution are verified', commercial.providerRevenueVerified === true, 'provider', 'Attach exact provider catalog and product-attributed revenue evidence.'),
+      ],
+    },
+  ];
+
+  for (const area of areas) {
+    const passed = area.checks.filter((row) => row.passed).length;
+    area.score = Number((10 * passed / area.checks.length).toFixed(1));
+    area.grade = area.score === 10 ? 'A+' : area.score >= 9 ? 'A' : area.score >= 8 ? 'B' : area.score >= 7 ? 'C' : area.score >= 6 ? 'D' : 'F';
+    area.status = area.score === 10 ? 'verified' : passed === 0 ? 'blocked' : 'partial';
+  }
+
+  const score = Number((areas.reduce((sum, area) => sum + area.score, 0) / areas.length).toFixed(1));
+  const atTarget = areas.every((area) => area.score === 10);
+  return {
+    scorecardVersion: SCORECARD_VERSION,
+    generatedAt: new Date().toISOString(),
+    target: { score: 10, grade: 'A+' },
+    atTarget,
+    score,
+    grade: atTarget ? 'A+' : score >= 9 ? 'A' : score >= 8 ? 'B' : score >= 7 ? 'C' : score >= 6 ? 'D' : 'F',
+    areas,
+    blockers: areas.flatMap((area) => area.checks
+      .filter((row) => !row.passed)
+      .map((row) => ({ area: area.id, check: row.id, evidenceClass: row.evidenceClass, remediation: row.remediation }))),
+  };
+}
+
+function formatMarkdown(report) {
+  const lines = [
+    '# ThumbGate A+ evidence scorecard',
+    '',
+    `Overall: **${report.score}/10 (${report.grade})**`,
+    `Target verified: **${report.atTarget ? 'YES' : 'NO'}**`,
+    '',
+    '| Area | Score | Grade | Status |',
+    '|---|---:|:---:|---|',
+    ...report.areas.map((area) => `| ${area.label} | ${area.score}/10 | ${area.grade} | ${area.status} |`),
+    '',
+    '## Remaining evidence blockers',
+    '',
+    ...(report.blockers.length
+      ? report.blockers.map((row) => `- **${row.area}/${row.check}** (${row.evidenceClass}): ${row.remediation}`)
+      : ['- None. Every repository, production, provider, security, and commercial check is verified.']),
+    '',
+  ];
+  return lines.join('\n');
+}
+
+function loadLiveEvidence(argv) {
+  const index = argv.indexOf('--evidence');
+  if (index === -1 || !argv[index + 1]) return {};
+  return JSON.parse(fs.readFileSync(path.resolve(argv[index + 1]), 'utf8'));
+}
+
+function main() {
+  const live = loadLiveEvidence(process.argv.slice(2));
+  const repo = collectRepositoryEvidence();
+  const report = evaluateReadiness({ repo, live });
+  process.stdout.write(process.argv.includes('--json')
+    ? `${JSON.stringify(report, null, 2)}\n`
+    : `${formatMarkdown(report)}\n`);
+  if (process.argv.includes('--require-a-plus') && !report.atTarget) process.exitCode = 1;
+}
+
+function isCliEntrypoint(argv = process.argv) {
+  return Boolean(argv[1]) && path.resolve(argv[1]) === path.resolve(__filename);
+}
+
+if (isCliEntrypoint()) main();
+
+module.exports = {
+  SCORECARD_VERSION,
+  collectRepositoryEvidence,
+  evaluateReadiness,
+  formatMarkdown,
+  isCliEntrypoint,
+};