chore(governance): expand ROUTE-002/003 holdout for R11 boundaries

Add approved-spec plan-only adversarial trap, dispatch routing probe,
and design-comparison holdout phrase. Bump fixture minimums (101/35)
with anti-drift tests; no router changes.

Author: johnteee

reflective-prompt-library/PROJECT_KNOWLEDGE.md

@@ -81,6 +81,7 @@ deferred promotions are recurrence-gated — see [panel backlog](plans/multi-age
 - 2026-06-18 Adopt STORM's source-grounded perspective discovery as an optional `reflective-research` method, not a new skill → [record](plans/storm-perspective-discovery-reflection-2026-06-18.md)
 - 2026-06-20 Synthesize external-adoption case studies, then correct the evidence and promotion-gate scope after the Test Plan routing counterexample → [record](plans/external-adoption-case-studies-2026-06-20.md)
 - 2026-06-21 Evaluate Hyperplan / multi-agent adversarial planning — no change (runtime = non-goal; methodology overlaps; possible gaps not promoted) → [record](plans/external-adoption-case-studies-2026-06-20.md)
+- 2026-06-25 Post-Round 68 holdout expansion — ROUTE-002 design comparison phrase; ROUTE-003 approved-spec plan-only trap + dispatch routing probe → [ROUTING_CONTRACT.md](plans/ROUTING_CONTRACT.md#r11-approved-spec-delivery)
 - 2026-06-25 Round 68 panel — full-doc drift sync, README/methodology-map anti-drift, reject holdout expansion at 100% → [record](plans/multi-agent-panel-consensus-2026-06-25.md)
 - 2026-06-25 Post-Round 68 maintenance — nine-skill wording parity (METHODOLOGY_MAP, skill-map, QUALITY_GATES), benchmark B024 dispatch (24 tasks, 9/9 workflows) → [QUALITY_GATES_SUMMARY.md](plans/QUALITY_GATES_SUMMARY.md)
 - 2026-06-25 Round 67 panel — CONTRIBUTING R8–R12 sync, ROUTING_CONTRACT related artifacts, contract anti-drift tests → [record](plans/multi-agent-panel-consensus-2026-06-25.md)

reflective-prompt-library/plans/QUALITY_GATES_SUMMARY.md

@@ -111,7 +111,7 @@ python3 reflective-prompt-library/plans/route_paraphrase_eval.py
 - Keeps a lower Phase-1 bar and separate aspirational target to avoid over-claiming
 
 **Results:**
-- Tested 36 holdout groups with 100 paraphrases
+- Tested 36 holdout groups with 101 paraphrases
 - Overall consistency: 100.0% (passes Phase-1 threshold >=80% and aspirational target >=90%)
 - Low-confidence route trace coverage: 100.0%
 - Review, clarification, workflow-design, workflow-implementation, research, and orchestration-selection phrasing now pass through boundary concepts
@@ -208,7 +208,7 @@ python3 reflective-prompt-library/plans/validate_skill_examples.py
 - Post-panel maintenance: ROUTING_CONTRACT **R11** approved-spec delivery (`implement_not_plan_trap`) at 100%
 
 **Results:**
-- Tested 13 adversarial groups with 31 paraphrases
+- Tested 14 adversarial groups with 35 paraphrases
 - Overall consistency: 100.0% (passes Phase-1 threshold >=80% and aspirational target >=90%)
 - Low-confidence route trace coverage: 100.0%
 - Output: `plans/route-003-results.json`
@@ -224,7 +224,7 @@ python3 reflective-prompt-library/plans/route_paraphrase_eval.py reflective-prom
 
 **What it does:**
 - Enforces minimum ROUTE-001/002/003 group and phrase counts before paraphrase eval runs
-- Current minimums: ROUTE-001 (12 intent + 4 adversarial groups, 128 phrases); ROUTE-002 (36 holdout groups, 100 phrases); ROUTE-003 (13 adversarial groups, 31 phrases)
+- Current minimums: ROUTE-001 (12 intent + 4 adversarial groups, 128 phrases); ROUTE-002 (36 holdout groups, 101 phrases); ROUTE-003 (14 adversarial groups, 35 phrases)
 - Round 22 panel compromise: deterministic hygiene without YAML dependency explosion
 - Integrated in `make validate` after skill examples gate; mirrored by pytest in `test_validate_route_fixture.py`
 - Cheatsheet parity anti-drift: `test_cheatsheet_boundary_quick_cues.py` (R12 quick-cue summary), `test_cheatsheet_*_parity.py` (holdout probe cues in EN/zh-TW cheatsheets)
@@ -279,8 +279,8 @@ The implementation aligns with research findings:
 | Skills with governance | 9/9 | ✅ Complete |
 | Benchmark tasks | 24 | ✅ Ready |
 | Routing consistency | 100.0% | ✅ Passes ROUTE-001 expanded boundary eval |
-| Holdout routing consistency | 100.0% | ✅ ROUTE-002 (36 groups, 100 paraphrases) |
-| Adversarial routing consistency | 100.0% | ✅ ROUTE-003 (13 groups, 31 paraphrases) |
+| Holdout routing consistency | 100.0% | ✅ ROUTE-002 (36 groups, 101 paraphrases) |
+| Adversarial routing consistency | 100.0% | ✅ ROUTE-003 (14 groups, 35 paraphrases) |
 | Skill example coverage | 9/9 | ✅ validate_skill_examples.py |
 | Linting errors | 0 | ✅ Clean |
 
@@ -311,7 +311,7 @@ ROUTE-002 measures unseen phrasing separately from ROUTE-001. Round 7 (2026-06-2
 ### Done ✅
 
 1. **CI/CD** — `.github/workflows/python-tools.yml` runs `make all` on push/PR
-2. **ROUTE-001/002/003 in CI** — 128 + 100 + 31 paraphrases at 100% consistency (seeded fixtures); `validate_route_fixture.py` gates minimum coverage
+2. **ROUTE-001/002/003 in CI** — 128 + 101 + 35 paraphrases at 100% consistency (seeded fixtures); `validate_route_fixture.py` gates minimum coverage
 3. **Governance validators** — links, lint, governance metadata, PROJECT_KNOWLEDGE, benchmark fixture, skill examples
 4. **Harness policy docs** — CONTRIBUTING, AGENTS, SKILL_INSTALLATION, maintenance playbook
 5. **Doc anti-drift** — `test_routing_contract.py`, cheatsheet parity tests, `test_readme_governance.py` (170+ pytest in CI)

reflective-prompt-library/plans/route-002-holdout-eval.yaml

@@ -362,6 +362,7 @@ holdout_sets:
       - compare two API designs on paper only no implementation
       - design comparison on paper without repo changes
       - 比較兩個 API 設計方案但不要寫 code
+      - compare API designs on paper without touching the repository
 
   - name: readme_plain_review_holdout
     expected_workflow: reflective-review

reflective-prompt-library/plans/route-003-adversarial-eval.yaml

@@ -124,6 +124,16 @@ adversarial_sets:
 
 
 
+  - name: approved_spec_plan_not_implement_trap
+    expected_workflow: reflective-spec-plan
+    expected_rigor: medium
+    expected_enhancements:
+      - acceptance_criteria
+    phrases:
+      - write tickets from the approved spec without implementing
+      - plan the approved spec without repo changes
+      - plan 已核准 spec without repo changes
+
   - name: dispatch_meta_skill_trap
     expected_workflow: reflective-dispatch
     expected_rigor: low
@@ -132,6 +142,7 @@ adversarial_sets:
     phrases:
       - which skill handles session handoff in this library
       - which reflective workflow skill covers handoff retro
+      - which reflective workflow skill fits a routing-only mixed intent
 
   - name: brief_not_plan_trap
     expected_workflow: reflective-brief

reflective-prompt-library/plans/tests/test_validate_route_fixture.py

@@ -109,6 +109,7 @@ def test_round_51_boundary_probes():
     ("lessons learned retro after this sprint", "reflective-handoff-retro"),
     ("compare official docs for both libraries before deciding", "reflective-research"),
     ("which reflective workflow skill covers handoff retro", "reflective-dispatch"),
+    ("plan the approved spec without repo changes", "reflective-spec-plan"),
     ("align stakeholders on goals before writing tickets", "reflective-brief"),
     ("釐清目標再拆工單", "reflective-brief"),
 )
@@ -207,3 +208,17 @@ def test_implement_approved_spec_not_plan_boundary():
         workflow, _, _, _ = router.route(text)
         assert workflow == expected, f"{text!r} -> {workflow}, want {expected}"
 
+def _approved_spec_plan_not_implement_trap_phrases() -> set[str]:
+    config = load_route_eval_config(PLANS / "route-003-adversarial-eval.yaml")
+    for group in config["adversarial_sets"]:
+        if group["name"] == "approved_spec_plan_not_implement_trap":
+            return set(group.get("phrases", []))
+    raise AssertionError("missing ROUTE-003 approved_spec_plan_not_implement_trap group")
+
+
+def test_approved_spec_plan_not_implement_trap_covers_probes():
+    """Anti-drift: ROUTE-003 plan-only approved-spec probes live in dedicated trap group."""
+    fixture_phrases = _approved_spec_plan_not_implement_trap_phrases()
+    missing = [p for p in IMPLEMENT_NOT_PLAN_SPEC_PLAN_PROBES if p not in fixture_phrases]
+    assert not missing, f"approved_spec_plan_not_implement_trap missing probes: {missing}"
+

reflective-prompt-library/plans/validate_route_fixture.py

@@ -29,10 +29,10 @@
 ROUTE_001_MIN_PHRASES = 128
 
 ROUTE_002_MIN_HOLDOUT_GROUPS = 36
-ROUTE_002_MIN_PHRASES = 100
+ROUTE_002_MIN_PHRASES = 101
 
-ROUTE_003_MIN_ADVERSARIAL_GROUPS = 13
-ROUTE_003_MIN_PHRASES = 31
+ROUTE_003_MIN_ADVERSARIAL_GROUPS = 14
+ROUTE_003_MIN_PHRASES = 35
 
 REQUIRED_TRACE_FIELDS = {
     "canonical_intent",