tribunal-kit 9.2.7 → 9.2.8

package/.agent/scripts/corpus_certification_orchestrator.js

@@ -0,0 +1,502 @@
+'use strict';
+
+const fs = require('fs');
+const path = require('path');
+const crypto = require('crypto');
+
+const { loadSkills } = require('./context_broker');
+const { parseContract } = require('./skill_contract_engine');
+const { classifySkill } = require('./skill_behavior_classifier');
+const { extractTestableClaims } = require('./skill_test_authoring_engine');
+const certEngine = require('./skill_certification_engine');
+const expansionEngine = require('./skill_coverage_expansion_engine');
+
+// ─── Constants ───────────────────────────────────────────────────────────────
+const ELIGIBILITY_CLASSES = {
+  EXECUTABLE: 'EXECUTABLE',
+  CONTRACT_EXECUTABLE: 'CONTRACT_EXECUTABLE',
+  COMPOSITE: 'COMPOSITE',
+  ROUTING: 'ROUTING',
+  GOVERNANCE: 'GOVERNANCE',
+  REFERENCE_ONLY: 'REFERENCE_ONLY',
+  HUMAN_REQUIRED: 'HUMAN_REQUIRED',
+  UNPROVABLE: 'UNPROVABLE',
+};
+
+const EXCLUDED_CLASSES = [
+  ELIGIBILITY_CLASSES.REFERENCE_ONLY,
+  ELIGIBILITY_CLASSES.HUMAN_REQUIRED,
+  ELIGIBILITY_CLASSES.UNPROVABLE,
+];
+
+const PILOT_SKILLS = [
+  'api-patterns',
+  'geo-fundamentals',
+  'hf-cloud-python-env-setup',
+  'lint-and-validate',
+  'webapp-testing',
+  'data-validation-schemas',
+  'config-validator',
+  'domain-modeling',
+  'react-specialist',
+  'sql-pro',
+  'backend-redis',
+  'devops-engineer',
+  'api-security-auditor',
+  'backend-security-expert',
+  'audit-and-fix',
+  'red-team-tactics',
+  'zero-trust-passkeys',
+  '60fps-animation',
+  'agent-organizer',
+  '12-principles-of-animation',
+  'apple-design',
+  'database-design',
+  'frontend-design',
+  'system-design-pro',
+  'taste-skill',
+];
+
+// ─── Section 6: Eligibility Classification ────────────────────────────────────
+function classifyCorpusEligibility(skillName, skillRaw = {}) {
+  const content = skillRaw.content || '';
+  const triggers = skillRaw.triggers || [];
+  const name = skillName.toLowerCase();
+
+  // 1. Human-required boundary check
+  if (
+    name.includes('taste') ||
+    name.includes('delight') ||
+    name.includes('whimsy') ||
+    name.includes('soft-skill')
+  ) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.HUMAN_REQUIRED,
+      human_boundary: true,
+      reason: 'Subjective human evaluation or aesthetic judgment boundary',
+    };
+  }
+
+  // 2. Reference-only check
+  const baseCls = classifySkill(skillName, skillRaw);
+  if (
+    baseCls.class === 'REFERENCE_ONLY' ||
+    name.includes('design') ||
+    name.includes('guidelines') ||
+    name.includes('principles') ||
+    name.includes('swiss-design') ||
+    name.includes('apple-design')
+  ) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.REFERENCE_ONLY,
+      human_boundary: false,
+      reason: 'Reference-only architecture or design guideline specification',
+    };
+  }
+
+  // 3. Routing / dispatch check
+  if (
+    name.includes('router') ||
+    name.includes('routing') ||
+    name.includes('orchestrat') ||
+    name.includes('dispatcher') ||
+    content.includes('router')
+  ) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.ROUTING,
+      human_boundary: false,
+      reason: 'Routing decision or agent dispatch layer',
+    };
+  }
+
+  // 4. Governance check
+  if (name.includes('governance') || name.includes('audit-and-fix')) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.GOVERNANCE,
+      human_boundary: false,
+      reason: 'Policy enforcement or static governance ruleset',
+    };
+  }
+
+  // 5. Composite check
+  if (skillRaw.dependencies && skillRaw.dependencies.length > 0) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.COMPOSITE,
+      human_boundary: false,
+      reason: 'Multi-skill composite orchestration dependency',
+    };
+  }
+
+  // 6. Contract-executable check
+  const { contract } = parseContract(content);
+  const hasInputs = contract?.inputs && Object.keys(contract.inputs).length > 0;
+  const hasOutputs = contract?.outputs && Object.keys(contract.outputs).length > 0;
+
+  if (
+    content.includes('npm run') ||
+    content.includes('node ') ||
+    content.includes('python ') ||
+    content.includes('bash ') ||
+    content.includes('scripts/') ||
+    name.includes('validator') ||
+    name.includes('pro') ||
+    name.includes('engineer')
+  ) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.EXECUTABLE,
+      human_boundary: false,
+      reason: 'Executable CLI or deterministic script runtime binding',
+    };
+  }
+
+  if (hasInputs && hasOutputs) {
+    return {
+      behavior_class: ELIGIBILITY_CLASSES.CONTRACT_EXECUTABLE,
+      human_boundary: false,
+      reason: 'Complete contract input and output schema specification',
+    };
+  }
+
+  return {
+    behavior_class: ELIGIBILITY_CLASSES.UNPROVABLE,
+    human_boundary: false,
+    reason: 'Insufficient executable or contract assertions to obtain evidence',
+  };
+}
+
+// ─── Section 11: Deterministic Priority Scoring ──────────────────────────────
+function calculatePriorityScore(skillRecord) {
+  let score = 0;
+  const breakdown = {};
+  const reasons = [];
+
+  // Security criticality (+40)
+  if (skillRecord.security_class) {
+    score += 40;
+    breakdown.security_criticality = 40;
+    reasons.push('+40 security criticality');
+  } else {
+    breakdown.security_criticality = 0;
+  }
+
+  // Executable evidence (+30)
+  if (skillRecord.behavior_class === ELIGIBILITY_CLASSES.EXECUTABLE) {
+    score += 30;
+    breakdown.executable_evidence = 30;
+    reasons.push('+30 executable evidence');
+  } else {
+    breakdown.executable_evidence = 0;
+  }
+
+  // Contract executable (+25)
+  if (
+    skillRecord.behavior_class === ELIGIBILITY_CLASSES.CONTRACT_EXECUTABLE ||
+    skillRecord.contract_status === 'VALID'
+  ) {
+    score += 25;
+    breakdown.contract_executable = 25;
+    reasons.push('+25 contract executable');
+  } else {
+    breakdown.contract_executable = 0;
+  }
+
+  // Existing external tests (+20)
+  if (skillRecord.has_external_tests) {
+    score += 20;
+    breakdown.existing_external_tests = 20;
+    reasons.push('+20 existing external tests');
+  } else {
+    breakdown.existing_external_tests = 0;
+  }
+
+  // High usage (+15)
+  if (skillRecord.routing_usage >= 5) {
+    score += 15;
+    breakdown.high_usage = 15;
+    reasons.push('+15 high usage');
+  } else {
+    breakdown.high_usage = 0;
+  }
+
+  // High routing importance (+10)
+  if (skillRecord.routing_importance >= 5) {
+    score += 10;
+    breakdown.high_routing_importance = 10;
+    reasons.push('+10 high routing importance');
+  } else {
+    breakdown.high_routing_importance = 0;
+  }
+
+  // Strong contract (+10)
+  if (skillRecord.strong_contract) {
+    score += 10;
+    breakdown.strong_contract = 10;
+    reasons.push('+10 strong contract');
+  } else {
+    breakdown.strong_contract = 0;
+  }
+
+  // Human boundary (-50)
+  if (skillRecord.human_boundary) {
+    score -= 50;
+    breakdown.human_boundary = -50;
+    reasons.push('-50 human boundary penalty');
+  } else {
+    breakdown.human_boundary = 0;
+  }
+
+  // Reference only (-100)
+  if (skillRecord.behavior_class === ELIGIBILITY_CLASSES.REFERENCE_ONLY) {
+    score -= 100;
+    breakdown.reference_only = -100;
+    reasons.push('-100 reference only penalty');
+  } else {
+    breakdown.reference_only = 0;
+  }
+
+  return {
+    raw_score: score,
+    breakdown,
+    reasons,
+    explanation: reasons.join('; '),
+  };
+}
+
+// ─── Section 5: Corpus Inventory ─────────────────────────────────────────────
+function buildCorpusInventory(options = {}) {
+  const agentDir = options.agentDir || path.resolve(__dirname, '..');
+  const skills = loadSkills(agentDir);
+  const repoRoot = path.resolve(agentDir, '..');
+  const p24CertPath = path.join(repoRoot, 'phase24', 'certification.json');
+
+  let p24Cert = { skills: {} };
+  if (fs.existsSync(p24CertPath)) {
+    try {
+      p24Cert = JSON.parse(fs.readFileSync(p24CertPath, 'utf8'));
+    } catch {}
+  }
+
+  const inventory = [];
+  const sorted = [...skills].sort((a, b) => a.name.localeCompare(b.name));
+
+  for (const s of sorted) {
+    const skillName = s.name;
+    const content = s.content || '';
+    const { contract } = parseContract(content);
+    const classification = classifyCorpusEligibility(skillName, s);
+
+    const isSecurity =
+      skillName.includes('security') ||
+      skillName.includes('red-team') ||
+      skillName.includes('vulnerability') ||
+      skillName.includes('audit-and-fix') ||
+      skillName.includes('zero-trust') ||
+      skillName.includes('injection') ||
+      skillName.includes('defense');
+
+    const claims = extractTestableClaims(skillName, s);
+    const p24Record = p24Cert.skills[skillName];
+    const currentCert = p24Record ? p24Record.status : 'UNPROVABLE';
+    const isPilot = PILOT_SKILLS.includes(skillName);
+
+    const hasStrongContract =
+      contract &&
+      contract.inputs &&
+      contract.outputs &&
+      Object.keys(contract.inputs).length > 0 &&
+      Object.keys(contract.outputs).length > 0;
+
+    const record = {
+      skill_id: skillName,
+      behavior_class: classification.behavior_class,
+      contract_status: contract ? 'VALID' : 'MISSING',
+      claim_count: claims.length,
+      testability: classification.behavior_class === ELIGIBILITY_CLASSES.REFERENCE_ONLY ? 'UNTESTABLE' : 'DIRECT',
+      current_certification: currentCert,
+      routing_usage: (s.triggers || []).length,
+      routing_importance: skillName.includes('pro') || skillName.includes('specialist') ? 6 : 2,
+      security_class: isSecurity,
+      human_boundary: classification.human_boundary,
+      strong_contract: !!hasStrongContract,
+      has_external_tests: fs.existsSync(path.join(agentDir, 'skills', skillName, 'tests')),
+      is_pilot: isPilot,
+      eligible: !isPilot && !EXCLUDED_CLASSES.includes(classification.behavior_class) && claims.length > 0,
+    };
+
+    const scoring = calculatePriorityScore(record);
+    record.priority_score = scoring.raw_score;
+    record.score_breakdown = scoring.breakdown;
+    record.score_explanation = scoring.explanation;
+
+    inventory.push(record);
+  }
+
+  return inventory;
+}
+
+// ─── Section 10 & 12: Wave Selection ─────────────────────────────────────────
+function selectWave(waveNumber = 1, inventory = [], options = {}) {
+  const maxWaveSize = options.maxWaveSize || 25;
+  const eligibleSkills = inventory.filter(s => s.eligible && s.current_certification === 'UNPROVABLE');
+
+  // Deterministic sort: descending priority score, tie-breaker: skill_id
+  const sorted = [...eligibleSkills].sort((a, b) => {
+    if (b.priority_score !== a.priority_score) {
+      return b.priority_score - a.priority_score;
+    }
+    return a.skill_id.localeCompare(b.skill_id);
+  });
+
+  const selected = sorted.slice(0, maxWaveSize);
+  const selectedIds = selected.map(s => s.skill_id);
+
+  const excluded = inventory.filter(s => !selectedIds.includes(s.skill_id));
+  const excludedRecords = excluded.map(s => ({
+    skill_id: s.skill_id,
+    reason: s.is_pilot
+      ? 'Already evaluated in pilot corpus'
+      : EXCLUDED_CLASSES.includes(s.behavior_class)
+      ? `Excluded class: ${s.behavior_class}`
+      : 'Deferred to later waves (capacity threshold)',
+    automated_certification: false,
+  }));
+
+  const selectionHash = crypto
+    .createHash('sha256')
+    .update(JSON.stringify(selectedIds))
+    .digest('hex');
+
+  return {
+    wave: waveNumber,
+    selected_skills: selectedIds,
+    selected_records: selected,
+    excluded_skills: excludedRecords,
+    selection_hash: selectionHash,
+    selection_method: 'DETERMINISTIC_CORPUS_PRIORITY',
+    timestamp: new Date().toISOString(),
+  };
+}
+
+// ─── Section 27-28: Wave Snapshots & Atomicity ───────────────────────────────
+function createWaveSnapshot(currentCertificationSnapshot, options = {}) {
+  return {
+    snapshot_hash: crypto.createHash('sha256').update(JSON.stringify(currentCertificationSnapshot)).digest('hex'),
+    timestamp: new Date().toISOString(),
+    skills: JSON.parse(JSON.stringify(currentCertificationSnapshot.skills || {})),
+    summary: JSON.parse(JSON.stringify(currentCertificationSnapshot.summary || {})),
+  };
+}
+
+function computeWaveDiff(preSnapshot, postSnapshot) {
+  return certEngine.compareCertificationSnapshots(preSnapshot, postSnapshot);
+}
+
+// ─── Section 27: Rollback ────────────────────────────────────────────────────
+function rollbackWave(waveNumber, preSnapshot, targetCertPath) {
+  fs.writeFileSync(targetCertPath, JSON.stringify(preSnapshot, null, 2), 'utf8');
+
+  // Section 24: Record WAVE_ROLLED_BACK to history ledger
+  const historyPath = path.resolve(__dirname, '..', 'data', 'corpus_certification_history.jsonl');
+  if (fs.existsSync(historyPath)) {
+    fs.appendFileSync(
+      historyPath,
+      JSON.stringify({
+        event: 'WAVE_ROLLED_BACK',
+        wave_id: waveNumber,
+        restored_hash: preSnapshot.snapshot_hash,
+        timestamp: new Date().toISOString(),
+      }) + '\n',
+      'utf8'
+    );
+  }
+
+  return {
+    wave: waveNumber,
+    status: 'ROLLED_BACK',
+    restored_hash: preSnapshot.snapshot_hash,
+    timestamp: new Date().toISOString(),
+  };
+}
+
+// ─── Section 30: Corpus Certification Metrics ────────────────────────────────
+function generateCorpusMetrics(inventory, certificationState, waveResults = []) {
+  const totalCorpus = inventory.length;
+  const skillsMap = certificationState.skills || {};
+
+  let certified = 0;
+  let highConf = 0;
+  let partial = 0;
+  let unprovable = 0;
+  let referenceOnly = 0;
+  let humanRequired = 0;
+  let securityCertified = 0;
+
+  for (const item of inventory) {
+    const cert = skillsMap[item.skill_id];
+    const status = cert ? cert.status : 'UNPROVABLE';
+
+    if (item.behavior_class === ELIGIBILITY_CLASSES.REFERENCE_ONLY) {
+      referenceOnly++;
+    }
+    if (item.human_boundary) {
+      humanRequired++;
+    }
+
+    if (status === certEngine.CERTIFICATION_LEVELS.BEHAVIORALLY_CERTIFIED) {
+      certified++;
+      if (item.security_class) securityCertified++;
+    } else if (status === certEngine.CERTIFICATION_LEVELS.HIGH_CONFIDENCE) {
+      highConf++;
+    } else if (status === certEngine.CERTIFICATION_LEVELS.PARTIAL) {
+      partial++;
+    } else {
+      unprovable++;
+    }
+  }
+
+  let totalDiscoveredClaims = 0;
+  let totalVerifiedClaims = 0;
+  let totalFailedClaims = 0;
+
+  for (const item of inventory) {
+    totalDiscoveredClaims += item.claim_count;
+    const cert = skillsMap[item.skill_id];
+    if (cert && cert.claims_summary) {
+      totalVerifiedClaims += cert.claims_summary.verified || 0;
+      totalFailedClaims += cert.claims_summary.failed || 0;
+    }
+  }
+
+  return {
+    total_corpus_skills: totalCorpus,
+    pilot_certified_skills: 18,
+    newly_certified_skills: certified - 18 > 0 ? certified - 18 : 0,
+    behaviorally_certified_skills: certified,
+    high_confidence_skills: highConf,
+    partial_skills: partial,
+    unprovable_skills: unprovable,
+    reference_only_skills: referenceOnly,
+    human_required_skills: humanRequired,
+    security_skills_certified: securityCertified,
+    total_claims_discovered: totalDiscoveredClaims,
+    total_claims_verified: totalVerifiedClaims,
+    total_claims_failed: totalFailedClaims,
+    total_evidence_promoted: waveResults.reduce((acc, w) => acc + (w.promoted_evidence || 0), 0),
+    waves_executed: waveResults.length,
+    timestamp: new Date().toISOString(),
+  };
+}
+
+module.exports = {
+  ELIGIBILITY_CLASSES,
+  EXCLUDED_CLASSES,
+  PILOT_SKILLS,
+  classifyCorpusEligibility,
+  calculatePriorityScore,
+  buildCorpusInventory,
+  selectWave,
+  createWaveSnapshot,
+  computeWaveDiff,
+  rollbackWave,
+  generateCorpusMetrics,
+};