diff --git a/CLAUDE.md b/CLAUDE.md
index 499c825..9c1c22e 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -2,7 +2,7 @@
 
 Voyage — a contract-driven Claude Code pipeline: brief, research, plan, execute, review, continue. Deep implementation planning and research with specialized agent swarms, external research, adversarial review, session decomposition, disciplined execution, and headless support.
 
-**Design principle: Context Engineering** — build the right context by orchestrating specialized agents. Each step in the pipeline (brief → research → plan → execute) produces a structured artifact that the next step consumes. (The claim that fanning out to a swarm *relieves* the main context is asserted-by-design, not measured end-to-end: the one delegation actually measured — the Phase-7 synthesis PoC — found Δ main-context ≈ 0, see `docs/T1-synthesis-poc-results.md`. The structured-artifact handoffs are the load-bearing benefit; main-context relief is not yet demonstrated.)
+**Design principle: Context Engineering** — build the right context by orchestrating specialized agents. Each step in the pipeline (brief → research → plan → execute) produces a structured artifact that the next step consumes. (The claim that fanning out to a swarm *relieves* the main context is asserted-by-design, not measured end-to-end: the one delegation actually measured — the Phase-7 synthesis PoC — found Δ main-context ≈ 0, see `docs/T1-synthesis-poc-results.md`. The parallel wall-clock + structured artifact handoffs are the load-bearing benefit; main-context relief is not yet demonstrated.)
 
 > **v3.0.0 — architect step extracted from this plugin.** The plan command still auto-discovers `architecture/overview.md` if present, so any compatible producer (architect plugin no longer publicly distributed; the architecture/overview.md slot remains available for any compatible producer) plugs into the same slot. See [CHANGELOG.md](CHANGELOG.md) for migration history.
 
diff --git a/tests/lib/doc-consistency.test.mjs b/tests/lib/doc-consistency.test.mjs
index 926923c..f8c03c9 100644
--- a/tests/lib/doc-consistency.test.mjs
+++ b/tests/lib/doc-consistency.test.mjs
@@ -1220,3 +1220,49 @@ test('S33: the analysis doc records the D1–D3 resolved forks (considered-and-k
   // reader sees the downgrade was reconsidered-and-declined, not overlooked.
   assert.ok(/40d8742/.test(t), 'decision record must cite the model-pin commit 40d8742 (D3 kept firm, document-only)');
 });
+
+// ── S38 — fan-out hedge harmonization + relief-assertion forward-guard (SC d) ──
+//
+// The brief's SC(d) premise — an unhedged "fan-out relieves main context" string
+// present in ~23 files awaiting a removal diff — is factually false: plan-time
+// enumeration shows the unhedged set is ∅. Every live relief mention is already
+// hedged (the sole one is CLAUDE.md:5, hedged since S18), and the
+// docs/T2-bakeoff-results.md "frees the main context" lines are a DIFFERENT, true
+// background-execution claim (must NOT be touched). SC(d) therefore ships as:
+//   (1) one surgical harmonization at CLAUDE.md:5 making the replacement phrase
+//       "parallel wall-clock" grep-positive (the Δ≈0 / "not yet demonstrated"
+//       qualifier kept intact), and
+//   (2) this durable forward-guard — a PRECISE relief-assertion matcher (verb-of-
+//       relief + main-context, NOT a bare "main context" grep that would false-fail
+//       benign locatives like "runs inline in main context") scoped to product-
+//       facing docs, requiring an adjacent hedge token on any match.
+// Fix the SOURCE prose (add the hedge / drop the claim), never this test.
+
+test('S38: CLAUDE.md carries the harmonized fan-out phrase "parallel wall-clock" (SC d)', () => {
+  assert.ok(
+    read('CLAUDE.md').includes('parallel wall-clock'),
+    'CLAUDE.md:5 must state the replacement phrase "parallel wall-clock + structured artifact handoffs" (SC d harmonization)',
+  );
+});
+
+test('S38: forward-guard — no product-facing doc asserts main-context relief unhedged (SC d)', () => {
+  // Banned form: a verb of relief adjacent to "main context" (or "main-context …
+  // relief"), NOT a plain locative ("runs in main context"). Decision/research
+  // records (docs/*) are append-only history and out of scope by design.
+  const RELIEF = /(relieve|relieves|relieving|free|frees|offload|offloads)\b[^.\n]{0,40}\bmain[-\s]context\b|\bmain[-\s]context\b[^.\n]{0,20}\brelief\b|swarm\s+relieves\s+context/i;
+  const HEDGE = /Δ|not measured|asserted-by-design|not yet demonstrated|not demonstrated/i;
+  const productFacing = ['README.md', 'CLAUDE.md', ...listMd('commands').map((f) => `commands/${f}`)];
+  const offenders = [];
+  for (const rel of productFacing) {
+    read(rel).split('\n').forEach((ln, i) => {
+      if (RELIEF.test(ln) && !HEDGE.test(ln)) offenders.push(`${rel}:${i + 1}  ${ln.trim().slice(0, 100)}`);
+    });
+  }
+  assert.equal(
+    offenders.length,
+    0,
+    'Unhedged main-context-relief assertion(s) in product-facing docs — add an adjacent hedge ' +
+      '(Δ / "not measured" / "asserted-by-design" / "not yet demonstrated") or drop the claim:\n' +
+      offenders.join('\n'),
+  );
+});