diff --git a/plugins/config-audit/scanners/lib/humanizer.mjs b/plugins/config-audit/scanners/lib/humanizer.mjs
new file mode 100644
index 0000000..62f1c46
--- /dev/null
+++ b/plugins/config-audit/scanners/lib/humanizer.mjs
@@ -0,0 +1,196 @@
+/**
+ * Plain-language humanizer for config-audit findings.
+ *
+ * Pure functions. Never mutate inputs. Translates technical scanner output
+ * into user-friendly language at output-formatting time. Adds three new
+ * fields to each finding:
+ *   - userImpactCategory: human-readable label per scanner (research/02)
+ *   - userActionLanguage: one-line urgency phrase per severity
+ *   - relevanceContext: deterministic file-pattern heuristic
+ *
+ * Original id, scanner, severity, file, line, evidence, category, autoFixable
+ * are preserved exactly. Title, description, recommendation are replaced when
+ * a translation is found; otherwise the originals are kept.
+ *
+ * Lookup order (per scanner prefix):
+ *   1. exact title in TRANSLATIONS[prefix].static
+ *   2. first regex match in TRANSLATIONS[prefix].patterns
+ *   3. TRANSLATIONS[prefix]._default
+ *   4. fallthrough: original strings (when scanner prefix has no entry)
+ *
+ * Zero external dependencies.
+ */
+
+import { TRANSLATIONS } from './humanizer-data.mjs';
+
+/**
+ * Map scanner prefix to user-facing impact-category label (research/02 line 124).
+ */
+const SCANNER_TO_CATEGORY = {
+  CML: 'Configuration mistake',
+  SET: 'Configuration mistake',
+  HKV: 'Configuration mistake',
+  RUL: 'Configuration mistake',
+  MCP: 'Configuration mistake',
+  IMP: 'Configuration mistake',
+  CNF: 'Conflict',
+  COL: 'Conflict',
+  TOK: 'Wasted tokens',
+  CPS: 'Wasted tokens',
+  DIS: 'Dead config',
+  GAP: 'Missed opportunity',
+  PLH: 'Configuration mistake',
+};
+
+/**
+ * Map severity to one-line action-language phrase (research/02 line 134).
+ */
+const SEVERITY_TO_ACTION = {
+  critical: 'Fix this now',
+  high: 'Fix soon',
+  medium: 'Fix when convenient',
+  low: 'Optional cleanup',
+  info: 'FYI',
+};
+
+/**
+ * Compute relevance context from a finding's file path. Deterministic, in-process,
+ * no subprocess. Conservative — defaults to 'affects-everyone' when ambiguous.
+ *
+ * @param {string|null|undefined} filePath
+ * @returns {'test-fixture-no-impact' | 'affects-this-machine-only' | 'affects-everyone'}
+ */
+export function computeRelevanceContext(filePath) {
+  if (typeof filePath !== 'string' || filePath.length === 0) {
+    return 'affects-everyone';
+  }
+  if (filePath.includes('/tests/fixtures/') || filePath.includes('/test/fixtures/')) {
+    return 'test-fixture-no-impact';
+  }
+  // Match basename pattern *.local.* (e.g., settings.local.json, claude.local.md)
+  const basename = filePath.split('/').pop() || '';
+  if (/\.local\./.test(basename)) {
+    return 'affects-this-machine-only';
+  }
+  return 'affects-everyone';
+}
+
+/**
+ * Look up translation for a finding by scanner prefix and title.
+ * Returns the translation object or null when no match (caller falls through to original).
+ *
+ * @param {string} scanner
+ * @param {string} title
+ * @returns {{title:string, description:string, recommendation:string} | null}
+ */
+function lookupTranslation(scanner, title) {
+  const entry = TRANSLATIONS[scanner];
+  if (!entry) return null;
+
+  // 1. Exact static match
+  if (typeof title === 'string' && entry.static && Object.prototype.hasOwnProperty.call(entry.static, title)) {
+    return entry.static[title];
+  }
+
+  // 2. Pattern match
+  if (Array.isArray(entry.patterns) && typeof title === 'string') {
+    for (const p of entry.patterns) {
+      if (p.regex instanceof RegExp && p.regex.test(title)) {
+        return p.translation;
+      }
+    }
+  }
+
+  // 3. Default
+  if (entry._default) {
+    return entry._default;
+  }
+
+  return null;
+}
+
+/**
+ * Humanize a single finding. Pure — never mutates input. Returns a new object.
+ *
+ * @param {object} finding - finding object from scanner output
+ * @returns {object} new finding with translated title/description/recommendation +
+ *                   userImpactCategory, userActionLanguage, relevanceContext fields
+ */
+export function humanizeFinding(finding) {
+  if (!finding || typeof finding !== 'object') {
+    return finding;
+  }
+
+  const translation = lookupTranslation(finding.scanner, finding.title);
+  const category = SCANNER_TO_CATEGORY[finding.scanner] || 'Other';
+  const action = SEVERITY_TO_ACTION[finding.severity] || 'FYI';
+  const relevance = computeRelevanceContext(finding.file);
+
+  const out = {
+    // Preserve identifying / structural fields exactly
+    id: finding.id,
+    scanner: finding.scanner,
+    severity: finding.severity,
+    // Replace prose if a translation exists; otherwise keep originals
+    title: translation ? translation.title : finding.title,
+    description: translation ? translation.description : finding.description,
+    file: finding.file ?? null,
+    line: finding.line ?? null,
+    evidence: finding.evidence ?? null,
+    category: finding.category ?? null,
+    recommendation: translation ? translation.recommendation : finding.recommendation,
+    autoFixable: finding.autoFixable ?? false,
+    // New humanized fields
+    userImpactCategory: category,
+    userActionLanguage: action,
+    relevanceContext: relevance,
+  };
+
+  // Preserve optional details payload if present (v5 N6)
+  if (finding.details && typeof finding.details === 'object') {
+    out.details = finding.details;
+  }
+
+  return out;
+}
+
+/**
+ * Humanize an array of findings. Pure — returns a new array of new objects.
+ *
+ * @param {object[]} findings
+ * @returns {object[]}
+ */
+export function humanizeFindings(findings) {
+  if (!Array.isArray(findings)) return findings;
+  return findings.map(humanizeFinding);
+}
+
+/**
+ * Humanize a top-level envelope produced by `runAllScanners`. Walks
+ * `env.scanners[].findings`. Pure — returns a new envelope with new
+ * scanner objects and new finding objects. The envelope-level shape
+ * (scanners array, target_path, total_duration_ms, aggregate, etc.)
+ * is preserved.
+ *
+ * @param {object} env
+ * @returns {object}
+ */
+export function humanizeEnvelope(env) {
+  if (!env || typeof env !== 'object' || !Array.isArray(env.scanners)) {
+    return env;
+  }
+
+  const newScanners = env.scanners.map((s) => {
+    if (!s || typeof s !== 'object') return s;
+    if (!Array.isArray(s.findings)) return s;
+    return {
+      ...s,
+      findings: humanizeFindings(s.findings),
+    };
+  });
+
+  return {
+    ...env,
+    scanners: newScanners,
+  };
+}
diff --git a/plugins/config-audit/tests/lib/humanizer.test.mjs b/plugins/config-audit/tests/lib/humanizer.test.mjs
new file mode 100644
index 0000000..a3f04e9
--- /dev/null
+++ b/plugins/config-audit/tests/lib/humanizer.test.mjs
@@ -0,0 +1,302 @@
+import { test } from 'node:test';
+import assert from 'node:assert/strict';
+import {
+  humanizeFinding,
+  humanizeFindings,
+  humanizeEnvelope,
+  computeRelevanceContext,
+} from '../../scanners/lib/humanizer.mjs';
+
+// ─── helpers ────────────────────────────────────────────────────────────
+
+function makeFinding(overrides = {}) {
+  return {
+    id: 'CA-CML-001',
+    scanner: 'CML',
+    severity: 'medium',
+    title: 'No CLAUDE.md found',
+    description: 'No CLAUDE.md file at the project root.',
+    file: '/Users/test/project/CLAUDE.md',
+    line: null,
+    evidence: 'evidence text',
+    category: 'config',
+    recommendation: 'Create one.',
+    autoFixable: false,
+    ...overrides,
+  };
+}
+
+// ─── purity ─────────────────────────────────────────────────────────────
+
+test('humanizeFinding does not mutate its input', () => {
+  const input = makeFinding();
+  const before = JSON.parse(JSON.stringify(input));
+  humanizeFinding(input);
+  assert.deepEqual(input, before, 'input was mutated');
+});
+
+test('humanizeFindings does not mutate its input array', () => {
+  const input = [makeFinding(), makeFinding({ id: 'CA-CML-002', title: 'Repeated content detected' })];
+  const before = JSON.parse(JSON.stringify(input));
+  humanizeFindings(input);
+  assert.deepEqual(input, before, 'input array was mutated');
+});
+
+test('humanizeEnvelope does not mutate its input', () => {
+  const env = {
+    target_path: '/tmp',
+    scanners: [
+      { scanner: 'CML', status: 'ok', findings: [makeFinding()], counts: {} },
+    ],
+  };
+  const before = JSON.parse(JSON.stringify(env));
+  humanizeEnvelope(env);
+  assert.deepEqual(env, before, 'envelope was mutated');
+});
+
+// ─── field preservation ────────────────────────────────────────────────
+
+test('humanizeFinding preserves id, scanner, severity, file, line, evidence, category, autoFixable', () => {
+  const input = makeFinding({
+    id: 'CA-CML-042',
+    scanner: 'CML',
+    severity: 'high',
+    file: '/tmp/x.md',
+    line: 17,
+    evidence: 'specific snippet',
+    category: 'cml',
+    autoFixable: true,
+  });
+  const out = humanizeFinding(input);
+  assert.equal(out.id, 'CA-CML-042');
+  assert.equal(out.scanner, 'CML');
+  assert.equal(out.severity, 'high');
+  assert.equal(out.file, '/tmp/x.md');
+  assert.equal(out.line, 17);
+  assert.equal(out.evidence, 'specific snippet');
+  assert.equal(out.category, 'cml');
+  assert.equal(out.autoFixable, true);
+});
+
+test('humanizeFinding preserves optional details payload', () => {
+  const input = makeFinding();
+  input.details = { foo: 'bar', count: 7 };
+  const out = humanizeFinding(input);
+  assert.deepEqual(out.details, { foo: 'bar', count: 7 });
+});
+
+// ─── translation lookup ────────────────────────────────────────────────
+
+test('humanizeFinding rewrites title for known static title', () => {
+  const input = makeFinding({ scanner: 'CML', title: 'No CLAUDE.md found' });
+  const out = humanizeFinding(input);
+  assert.notEqual(out.title, input.title, 'title should be translated');
+  assert.ok(out.title.toLowerCase().includes('instructions') || out.title.toLowerCase().includes('claude'),
+    `humanized title should mention instructions or claude, got: ${out.title}`);
+});
+
+test('humanizeFinding falls back to _default when title unknown', () => {
+  const input = makeFinding({ scanner: 'CML', title: 'Unrecognized brand-new finding title' });
+  const out = humanizeFinding(input);
+  assert.notEqual(out.title, input.title, '_default should kick in');
+  // CML _default mentions "instructions file"
+  assert.ok(/instructions file/i.test(out.title), `expected CML _default title, got: ${out.title}`);
+});
+
+test('humanizeFinding passes through original strings when scanner prefix unknown', () => {
+  const input = makeFinding({ scanner: 'XXX', title: 'whatever' });
+  const out = humanizeFinding(input);
+  assert.equal(out.title, 'whatever');
+  assert.equal(out.description, input.description);
+  assert.equal(out.recommendation, input.recommendation);
+});
+
+test('humanizeFinding matches pattern entries (template-literal titles)', () => {
+  const input = makeFinding({
+    scanner: 'COL',
+    title: 'Skill name "okr-helper" used by multiple plugins',
+  });
+  const out = humanizeFinding(input);
+  assert.notEqual(out.title, input.title);
+  assert.ok(/two plugins|same name|multiple/i.test(out.title) ||
+            /two plugins|same name|multiple/i.test(out.description),
+    `expected pattern match for COL multiple-plugins case, got title: ${out.title}`);
+});
+
+// ─── userActionLanguage ────────────────────────────────────────────────
+
+test('humanizeFinding maps severity=critical -> "Fix this now"', () => {
+  const out = humanizeFinding(makeFinding({ severity: 'critical' }));
+  assert.equal(out.userActionLanguage, 'Fix this now');
+});
+
+test('humanizeFinding maps severity=high -> "Fix soon"', () => {
+  const out = humanizeFinding(makeFinding({ severity: 'high' }));
+  assert.equal(out.userActionLanguage, 'Fix soon');
+});
+
+test('humanizeFinding maps severity=medium -> "Fix when convenient"', () => {
+  const out = humanizeFinding(makeFinding({ severity: 'medium' }));
+  assert.equal(out.userActionLanguage, 'Fix when convenient');
+});
+
+test('humanizeFinding maps severity=low -> "Optional cleanup"', () => {
+  const out = humanizeFinding(makeFinding({ severity: 'low' }));
+  assert.equal(out.userActionLanguage, 'Optional cleanup');
+});
+
+test('humanizeFinding maps severity=info -> "FYI"', () => {
+  const out = humanizeFinding(makeFinding({ severity: 'info' }));
+  assert.equal(out.userActionLanguage, 'FYI');
+});
+
+test('humanizeFinding falls back to "FYI" for unknown severity', () => {
+  const out = humanizeFinding(makeFinding({ severity: 'mystery' }));
+  assert.equal(out.userActionLanguage, 'FYI');
+});
+
+// ─── userImpactCategory ────────────────────────────────────────────────
+
+test('humanizeFinding sets category Configuration mistake for CML/SET/HKV/RUL/MCP/IMP/PLH', () => {
+  for (const s of ['CML', 'SET', 'HKV', 'RUL', 'MCP', 'IMP', 'PLH']) {
+    const out = humanizeFinding(makeFinding({ scanner: s }));
+    assert.equal(out.userImpactCategory, 'Configuration mistake', `${s} should map to Configuration mistake`);
+  }
+});
+
+test('humanizeFinding sets category Conflict for CNF/COL', () => {
+  for (const s of ['CNF', 'COL']) {
+    const out = humanizeFinding(makeFinding({ scanner: s }));
+    assert.equal(out.userImpactCategory, 'Conflict');
+  }
+});
+
+test('humanizeFinding sets category Wasted tokens for TOK/CPS', () => {
+  for (const s of ['TOK', 'CPS']) {
+    const out = humanizeFinding(makeFinding({ scanner: s }));
+    assert.equal(out.userImpactCategory, 'Wasted tokens');
+  }
+});
+
+test('humanizeFinding sets category Dead config for DIS', () => {
+  const out = humanizeFinding(makeFinding({ scanner: 'DIS' }));
+  assert.equal(out.userImpactCategory, 'Dead config');
+});
+
+test('humanizeFinding sets category Missed opportunity for GAP', () => {
+  const out = humanizeFinding(makeFinding({ scanner: 'GAP', title: 'No CLAUDE.md file' }));
+  assert.equal(out.userImpactCategory, 'Missed opportunity');
+});
+
+test('humanizeFinding sets category Other for unknown scanner', () => {
+  const out = humanizeFinding(makeFinding({ scanner: 'XXX' }));
+  assert.equal(out.userImpactCategory, 'Other');
+});
+
+// ─── relevanceContext ──────────────────────────────────────────────────
+
+test('computeRelevanceContext detects test-fixture paths', () => {
+  assert.equal(computeRelevanceContext('/repo/tests/fixtures/foo/CLAUDE.md'), 'test-fixture-no-impact');
+  assert.equal(computeRelevanceContext('/repo/test/fixtures/bar.json'), 'test-fixture-no-impact');
+});
+
+test('computeRelevanceContext detects local-only paths via .local. infix', () => {
+  assert.equal(computeRelevanceContext('/repo/.claude/settings.local.json'), 'affects-this-machine-only');
+  assert.equal(computeRelevanceContext('/repo/CLAUDE.local.md'), 'affects-this-machine-only');
+});
+
+test('computeRelevanceContext defaults to affects-everyone for normal paths', () => {
+  assert.equal(computeRelevanceContext('/repo/CLAUDE.md'), 'affects-everyone');
+  assert.equal(computeRelevanceContext('/repo/.claude/settings.json'), 'affects-everyone');
+});
+
+test('computeRelevanceContext defaults to affects-everyone for null/empty paths', () => {
+  assert.equal(computeRelevanceContext(null), 'affects-everyone');
+  assert.equal(computeRelevanceContext(undefined), 'affects-everyone');
+  assert.equal(computeRelevanceContext(''), 'affects-everyone');
+});
+
+test('humanizeFinding sets relevanceContext from file', () => {
+  const f = makeFinding({ file: '/repo/tests/fixtures/x.json' });
+  assert.equal(humanizeFinding(f).relevanceContext, 'test-fixture-no-impact');
+
+  const g = makeFinding({ file: '/repo/.claude/settings.local.json' });
+  assert.equal(humanizeFinding(g).relevanceContext, 'affects-this-machine-only');
+
+  const h = makeFinding({ file: '/repo/CLAUDE.md' });
+  assert.equal(humanizeFinding(h).relevanceContext, 'affects-everyone');
+});
+
+// ─── humanizeFindings & humanizeEnvelope ──────────────────────────────
+
+test('humanizeFindings translates each finding in the array', () => {
+  const findings = [
+    makeFinding({ scanner: 'CML', title: 'No CLAUDE.md found' }),
+    makeFinding({ id: 'CA-CML-002', scanner: 'CML', title: 'Uses HTML comments' }),
+  ];
+  const out = humanizeFindings(findings);
+  assert.equal(out.length, 2);
+  assert.notEqual(out[0].title, findings[0].title);
+  assert.notEqual(out[1].title, findings[1].title);
+});
+
+test('humanizeFindings returns input unchanged if not an array', () => {
+  assert.equal(humanizeFindings(null), null);
+  assert.equal(humanizeFindings(undefined), undefined);
+});
+
+test('humanizeEnvelope walks scanners[].findings and humanizes each', () => {
+  const env = {
+    target_path: '/tmp',
+    scanners: [
+      {
+        scanner: 'CML',
+        status: 'ok',
+        findings: [makeFinding({ scanner: 'CML', title: 'No CLAUDE.md found' })],
+        counts: {},
+      },
+      {
+        scanner: 'TOK',
+        status: 'ok',
+        findings: [makeFinding({
+          id: 'CA-TOK-001',
+          scanner: 'TOK',
+          severity: 'low',
+          title: 'Cache-breaking volatile content at top of CLAUDE.md',
+          file: '/tmp/CLAUDE.md',
+        })],
+        counts: {},
+      },
+    ],
+  };
+  const out = humanizeEnvelope(env);
+  assert.equal(out.target_path, '/tmp');
+  assert.equal(out.scanners.length, 2);
+  assert.notEqual(out.scanners[0].findings[0].title, env.scanners[0].findings[0].title);
+  assert.notEqual(out.scanners[1].findings[0].title, env.scanners[1].findings[0].title);
+  assert.equal(out.scanners[1].findings[0].userImpactCategory, 'Wasted tokens');
+});
+
+test('humanizeEnvelope returns input unchanged if shape is wrong', () => {
+  assert.equal(humanizeEnvelope(null), null);
+  assert.equal(humanizeEnvelope({}).scanners, undefined);  // unchanged object
+  assert.equal(humanizeEnvelope({ scanners: 'not-an-array' }).scanners, 'not-an-array');
+});
+
+// ─── new fields presence ───────────────────────────────────────────────
+
+test('humanizeFinding always sets the three new fields', () => {
+  const out = humanizeFinding(makeFinding());
+  assert.equal(typeof out.userImpactCategory, 'string');
+  assert.equal(typeof out.userActionLanguage, 'string');
+  assert.equal(typeof out.relevanceContext, 'string');
+});
+
+// ─── ANSI-free guarantee ───────────────────────────────────────────────
+
+test('humanized output contains no ANSI escape sequences', () => {
+  const out = humanizeFinding(makeFinding({ scanner: 'CML', title: 'No CLAUDE.md found' }));
+  const allText = `${out.title} ${out.description} ${out.recommendation}`;
+  // eslint-disable-next-line no-control-regex
+  assert.equal(/\[/.test(allText), false, 'ANSI escape detected in humanized output');
+});