diff --git a/CLAUDE.md b/CLAUDE.md
index 82df6ce..8d8eceb 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -162,14 +162,14 @@ Implemented in `src/core/piece/evaluation/RuleEvaluator.ts`. The matched method
   3. User request (`{task}` — auto-injected unless placeholder present)
   4. Previous response (auto-injected if `pass_previous_response: true`)
   5. User inputs (auto-injected unless `{user_inputs}` placeholder present)
-  6. `instruction_template` content
+  6. `instruction` content
   7. Status output rules (auto-injected for tag-based rules)
 - Localized for `en` and `ja`
 - Related: `ReportInstructionBuilder` (Phase 2), `StatusJudgmentBuilder` (Phase 3)
 
 **Agent Runner** (`src/agents/runner.ts`)
 - Resolves agent specs (name or path) to agent configurations
-- Agent is optional — movements can execute with `instruction_template` only (no system prompt)
+- Agent is optional — movements can execute with `instruction` only (no system prompt)
 - 5-layer resolution for provider/model: CLI `--provider` / `--model` → persona_providers → movement override → project `.takt/config.yaml` → global `~/.takt/config.yaml`
 - Custom personas via `~/.takt/personas/<name>.md` or prompt files (.md)
 - Inline system prompts: If agent file doesn't exist, the agent string is used as inline system prompt
@@ -299,7 +299,7 @@ loop_monitors:
     threshold: 3                # Cycles before triggering judge
     judge:
       persona: supervisor
-      instruction_template: "Evaluate if the fix loop is making progress..."
+      instruction: "Evaluate if the fix loop is making progress..."
       rules:
         - condition: "Progress is being made"
           next: fix
@@ -342,7 +342,7 @@ movements:
       my-server:
         command: npx
         args: [-y, my-mcp-server]
-    instruction_template: |
+    instruction: |
       Custom instructions for this movement.
       {task}, {previous_response} are auto-injected if not present as placeholders.
     pass_previous_response: true        # Default: true
@@ -413,7 +413,7 @@ movements:
       part_edit: true                  # Edit permission for parts
       part_permission_mode: edit       # Permission mode for parts
       part_allowed_tools: [Read, Glob, Grep, Edit, Write, Bash]
-    instruction_template: |
+    instruction: |
       Decompose this task into independent subtasks.
     rules:
       - condition: "All parts completed"
@@ -549,7 +549,7 @@ Key rules:
 - Policy REJECT lists are what reviewers enforce. If a criterion is not in the policy REJECT list, reviewers will not catch it — even if knowledge explains the reasoning
 - Knowledge provides the WHY behind policy criteria. Knowledge alone does not trigger enforcement
 - Instructions are bound to a single piece movement. They reference procedures, not principles
-- Piece YAML `instruction_template` is for movement-specific details (which reports to read, movement routing, output templates)
+- Piece YAML `instruction` is for movement-specific details (which reports to read, movement routing, output templates)
 
 **Separation of concerns in piece engine:**
 - `PieceEngine` - Orchestration, state management, event emission
diff --git a/builtins/en/pieces/backend-cqrs-mini.yaml b/builtins/en/pieces/backend-cqrs-mini.yaml
index 0073532..c2cf579 100644
--- a/builtins/en/pieces/backend-cqrs-mini.yaml
+++ b/builtins/en/pieces/backend-cqrs-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: fix_both
diff --git a/builtins/en/pieces/backend-cqrs-review-fix.yaml b/builtins/en/pieces/backend-cqrs-review-fix.yaml
index c7232b0..2e68b48 100644
--- a/builtins/en/pieces/backend-cqrs-review-fix.yaml
+++ b/builtins/en/pieces/backend-cqrs-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/backend-cqrs.yaml b/builtins/en/pieces/backend-cqrs.yaml
index 61031dd..4c02c46 100644
--- a/builtins/en/pieces/backend-cqrs.yaml
+++ b/builtins/en/pieces/backend-cqrs.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/backend-mini.yaml b/builtins/en/pieces/backend-mini.yaml
index c1b2d48..6487056 100644
--- a/builtins/en/pieces/backend-mini.yaml
+++ b/builtins/en/pieces/backend-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: fix_both
diff --git a/builtins/en/pieces/backend-review-fix.yaml b/builtins/en/pieces/backend-review-fix.yaml
index 382c536..9b0f9de 100644
--- a/builtins/en/pieces/backend-review-fix.yaml
+++ b/builtins/en/pieces/backend-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/backend.yaml b/builtins/en/pieces/backend.yaml
index 490cf07..f72b893 100644
--- a/builtins/en/pieces/backend.yaml
+++ b/builtins/en/pieces/backend.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/compound-eye.yaml b/builtins/en/pieces/compound-eye.yaml
index 33ca603..db949dd 100644
--- a/builtins/en/pieces/compound-eye.yaml
+++ b/builtins/en/pieces/compound-eye.yaml
@@ -108,7 +108,7 @@ movements:
     rules:
       - condition: synthesis complete
         next: COMPLETE
-    instruction_template: |
+    instruction: |
       Two models (Claude / Codex) independently answered the same instruction.
       Synthesize their responses.
 
diff --git a/builtins/en/pieces/default.yaml b/builtins/en/pieces/default.yaml
index ce5da73..ef3cc82 100644
--- a/builtins/en/pieces/default.yaml
+++ b/builtins/en/pieces/default.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The ai_review ↔ ai_fix loop has repeated {cycle_count} times.
 
         Review the reports from each cycle and determine whether this loop
@@ -39,7 +39,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/dual-cqrs-mini.yaml b/builtins/en/pieces/dual-cqrs-mini.yaml
index 6142468..39c4bcb 100644
--- a/builtins/en/pieces/dual-cqrs-mini.yaml
+++ b/builtins/en/pieces/dual-cqrs-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: fix_both
diff --git a/builtins/en/pieces/dual-cqrs-review-fix.yaml b/builtins/en/pieces/dual-cqrs-review-fix.yaml
index b0c2820..3e98dc8 100644
--- a/builtins/en/pieces/dual-cqrs-review-fix.yaml
+++ b/builtins/en/pieces/dual-cqrs-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/dual-cqrs.yaml b/builtins/en/pieces/dual-cqrs.yaml
index a9af202..ff92413 100644
--- a/builtins/en/pieces/dual-cqrs.yaml
+++ b/builtins/en/pieces/dual-cqrs.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-ai-fix
+      instruction: loop-monitor-ai-fix
       rules:
         - condition: Healthy (making progress)
           next: ai_review
@@ -27,7 +27,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/dual-mini.yaml b/builtins/en/pieces/dual-mini.yaml
index 8fc544d..78ae752 100644
--- a/builtins/en/pieces/dual-mini.yaml
+++ b/builtins/en/pieces/dual-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: fix_both
diff --git a/builtins/en/pieces/dual-review-fix.yaml b/builtins/en/pieces/dual-review-fix.yaml
index 3d468d1..bed0afe 100644
--- a/builtins/en/pieces/dual-review-fix.yaml
+++ b/builtins/en/pieces/dual-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/dual.yaml b/builtins/en/pieces/dual.yaml
index 2959bae..e291a99 100644
--- a/builtins/en/pieces/dual.yaml
+++ b/builtins/en/pieces/dual.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-ai-fix
+      instruction: loop-monitor-ai-fix
       rules:
         - condition: Healthy (making progress)
           next: ai_review
@@ -27,7 +27,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/e2e-test.yaml b/builtins/en/pieces/e2e-test.yaml
index 7648ab5..350b393 100644
--- a/builtins/en/pieces/e2e-test.yaml
+++ b/builtins/en/pieces/e2e-test.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The ai_review ↔ ai_fix loop has repeated {cycle_count} times.
 
         Review the reports from each cycle and determine whether this loop
diff --git a/builtins/en/pieces/frontend-mini.yaml b/builtins/en/pieces/frontend-mini.yaml
index f4824cb..40ed9dd 100644
--- a/builtins/en/pieces/frontend-mini.yaml
+++ b/builtins/en/pieces/frontend-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: fix_both
diff --git a/builtins/en/pieces/frontend-review-fix.yaml b/builtins/en/pieces/frontend-review-fix.yaml
index c42b36d..ebbff53 100644
--- a/builtins/en/pieces/frontend-review-fix.yaml
+++ b/builtins/en/pieces/frontend-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/frontend.yaml b/builtins/en/pieces/frontend.yaml
index 9a23b56..66bda37 100644
--- a/builtins/en/pieces/frontend.yaml
+++ b/builtins/en/pieces/frontend.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: converging (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/magi.yaml b/builtins/en/pieces/magi.yaml
index f13ece5..595e391 100644
--- a/builtins/en/pieces/magi.yaml
+++ b/builtins/en/pieces/magi.yaml
@@ -18,7 +18,7 @@ initial_movement: melchior
 movements:
   - name: melchior
     persona: melchior
-    instruction_template: |
+    instruction: |
       # MAGI System Initiated
 
       ## Matter for Deliberation
@@ -48,7 +48,7 @@ movements:
         next: balthasar
   - name: balthasar
     persona: balthasar
-    instruction_template: |
+    instruction: |
       # MAGI System Continuing
 
       ## Matter for Deliberation
@@ -82,7 +82,7 @@ movements:
         next: casper
   - name: casper
     persona: casper
-    instruction_template: |
+    instruction: |
       # MAGI System Final Deliberation
 
       ## Matter for Deliberation
diff --git a/builtins/en/pieces/review-fix.yaml b/builtins/en/pieces/review-fix.yaml
index ed91098..643c8ab 100644
--- a/builtins/en/pieces/review-fix.yaml
+++ b/builtins/en/pieces/review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/review.yaml b/builtins/en/pieces/review.yaml
index ff15cc8..133d43b 100644
--- a/builtins/en/pieces/review.yaml
+++ b/builtins/en/pieces/review.yaml
@@ -177,7 +177,7 @@ movements:
     rules:
       - condition: Review synthesis complete
         next: COMPLETE
-    instruction_template: |
+    instruction: |
       ## Review Results
       {previous_response}
 
diff --git a/builtins/en/pieces/takt-default-review-fix.yaml b/builtins/en/pieces/takt-default-review-fix.yaml
index 7c6b333..f14ec6a 100644
--- a/builtins/en/pieces/takt-default-review-fix.yaml
+++ b/builtins/en/pieces/takt-default-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (progress being made)
           next: reviewers
diff --git a/builtins/en/pieces/takt-default-team-leader.yaml b/builtins/en/pieces/takt-default-team-leader.yaml
index de9bbbd..e9f06a3 100644
--- a/builtins/en/pieces/takt-default-team-leader.yaml
+++ b/builtins/en/pieces/takt-default-team-leader.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-ai-fix
+      instruction: loop-monitor-ai-fix
       rules:
         - condition: Healthy (making progress)
           next: ai_review
@@ -27,7 +27,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/takt-default.yaml b/builtins/en/pieces/takt-default.yaml
index 8687c5e..39791bd 100644
--- a/builtins/en/pieces/takt-default.yaml
+++ b/builtins/en/pieces/takt-default.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The ai_review ↔ ai_fix loop has repeated {cycle_count} times.
 
         Review the reports from each cycle and determine whether this loop
@@ -39,7 +39,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: Healthy (findings decreasing, fixes applied)
           next: reviewers
diff --git a/builtins/en/pieces/terraform.yaml b/builtins/en/pieces/terraform.yaml
index 6523f03..6f34cb9 100644
--- a/builtins/en/pieces/terraform.yaml
+++ b/builtins/en/pieces/terraform.yaml
@@ -265,7 +265,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The review → fix cycle has repeated {cycle_count} times.
         Check the review report history in the Report Directory and assess convergence.
 
@@ -282,7 +282,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The AI fix → review cycle has repeated {cycle_count} times.
         Check the review report history in the Report Directory and assess convergence.
 
@@ -299,7 +299,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The supervisor fix → review cycle has repeated {cycle_count} times.
         Check the review report history in the Report Directory and assess convergence.
 
diff --git a/builtins/en/pieces/unit-test.yaml b/builtins/en/pieces/unit-test.yaml
index b6cb47a..45243eb 100644
--- a/builtins/en/pieces/unit-test.yaml
+++ b/builtins/en/pieces/unit-test.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         The ai_review ↔ ai_fix loop has repeated {cycle_count} times.
 
         Review the reports from each cycle and determine whether this loop
diff --git a/builtins/ja/pieces/backend-cqrs-mini.yaml b/builtins/ja/pieces/backend-cqrs-mini.yaml
index 5498706..a1540a6 100644
--- a/builtins/ja/pieces/backend-cqrs-mini.yaml
+++ b/builtins/ja/pieces/backend-cqrs-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: fix_both
diff --git a/builtins/ja/pieces/backend-cqrs-review-fix.yaml b/builtins/ja/pieces/backend-cqrs-review-fix.yaml
index 9c79494..d25e94b 100644
--- a/builtins/ja/pieces/backend-cqrs-review-fix.yaml
+++ b/builtins/ja/pieces/backend-cqrs-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/backend-cqrs.yaml b/builtins/ja/pieces/backend-cqrs.yaml
index 5f6587a..bc70fcf 100644
--- a/builtins/ja/pieces/backend-cqrs.yaml
+++ b/builtins/ja/pieces/backend-cqrs.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/backend-mini.yaml b/builtins/ja/pieces/backend-mini.yaml
index eb4170c..15599ed 100644
--- a/builtins/ja/pieces/backend-mini.yaml
+++ b/builtins/ja/pieces/backend-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: fix_both
diff --git a/builtins/ja/pieces/backend-review-fix.yaml b/builtins/ja/pieces/backend-review-fix.yaml
index 31bdf18..6d14ca8 100644
--- a/builtins/ja/pieces/backend-review-fix.yaml
+++ b/builtins/ja/pieces/backend-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/backend.yaml b/builtins/ja/pieces/backend.yaml
index ab3c58b..f3d771f 100644
--- a/builtins/ja/pieces/backend.yaml
+++ b/builtins/ja/pieces/backend.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/compound-eye.yaml b/builtins/ja/pieces/compound-eye.yaml
index 245c8a7..6c0da93 100644
--- a/builtins/ja/pieces/compound-eye.yaml
+++ b/builtins/ja/pieces/compound-eye.yaml
@@ -106,7 +106,7 @@ movements:
     rules:
       - condition: 統合完了
         next: COMPLETE
-    instruction_template: |
+    instruction: |
       2つのモデル（Claude / Codex）が同じ指示に対して独立に回答しました。
       両者の回答を統合してください。
 
diff --git a/builtins/ja/pieces/default.yaml b/builtins/ja/pieces/default.yaml
index 71f0280..3457f6e 100644
--- a/builtins/ja/pieces/default.yaml
+++ b/builtins/ja/pieces/default.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         ai_review と ai_fix のループが {cycle_count} 回繰り返されました。
 
         各サイクルのレポートを確認し、このループが健全（進捗がある）か、
@@ -39,7 +39,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/dual-cqrs-mini.yaml b/builtins/ja/pieces/dual-cqrs-mini.yaml
index 4ce2652..abffd2a 100644
--- a/builtins/ja/pieces/dual-cqrs-mini.yaml
+++ b/builtins/ja/pieces/dual-cqrs-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: fix_both
diff --git a/builtins/ja/pieces/dual-cqrs-review-fix.yaml b/builtins/ja/pieces/dual-cqrs-review-fix.yaml
index c6b213f..9f2e887 100644
--- a/builtins/ja/pieces/dual-cqrs-review-fix.yaml
+++ b/builtins/ja/pieces/dual-cqrs-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/dual-cqrs.yaml b/builtins/ja/pieces/dual-cqrs.yaml
index ddae703..83149ff 100644
--- a/builtins/ja/pieces/dual-cqrs.yaml
+++ b/builtins/ja/pieces/dual-cqrs.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-ai-fix
+      instruction: loop-monitor-ai-fix
       rules:
         - condition: 健全（進捗あり）
           next: ai_review
@@ -27,7 +27,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/dual-mini.yaml b/builtins/ja/pieces/dual-mini.yaml
index f81c7ff..d77386d 100644
--- a/builtins/ja/pieces/dual-mini.yaml
+++ b/builtins/ja/pieces/dual-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: fix_both
diff --git a/builtins/ja/pieces/dual-review-fix.yaml b/builtins/ja/pieces/dual-review-fix.yaml
index 51b19b3..268e949 100644
--- a/builtins/ja/pieces/dual-review-fix.yaml
+++ b/builtins/ja/pieces/dual-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/dual.yaml b/builtins/ja/pieces/dual.yaml
index 1b5f7dc..38b0bdb 100644
--- a/builtins/ja/pieces/dual.yaml
+++ b/builtins/ja/pieces/dual.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-ai-fix
+      instruction: loop-monitor-ai-fix
       rules:
         - condition: 健全（進捗あり）
           next: ai_review
@@ -27,7 +27,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/e2e-test.yaml b/builtins/ja/pieces/e2e-test.yaml
index e3f76ed..cda5dc1 100644
--- a/builtins/ja/pieces/e2e-test.yaml
+++ b/builtins/ja/pieces/e2e-test.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         ai_review と ai_fix のループが {cycle_count} 回繰り返されました。
 
         各サイクルのレポートを確認し、このループが健全（進捗がある）か、
diff --git a/builtins/ja/pieces/frontend-mini.yaml b/builtins/ja/pieces/frontend-mini.yaml
index f7050f7..4554460 100644
--- a/builtins/ja/pieces/frontend-mini.yaml
+++ b/builtins/ja/pieces/frontend-mini.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: fix_both
diff --git a/builtins/ja/pieces/frontend-review-fix.yaml b/builtins/ja/pieces/frontend-review-fix.yaml
index c45d73c..551264b 100644
--- a/builtins/ja/pieces/frontend-review-fix.yaml
+++ b/builtins/ja/pieces/frontend-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/frontend.yaml b/builtins/ja/pieces/frontend.yaml
index f96a251..7af1b0e 100644
--- a/builtins/ja/pieces/frontend.yaml
+++ b/builtins/ja/pieces/frontend.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/magi.yaml b/builtins/ja/pieces/magi.yaml
index e0717e5..d09558d 100644
--- a/builtins/ja/pieces/magi.yaml
+++ b/builtins/ja/pieces/magi.yaml
@@ -18,7 +18,7 @@ initial_movement: melchior
 movements:
   - name: melchior
     persona: melchior
-    instruction_template: |
+    instruction: |
       # MAGI System 起動
 
       ## 審議事項
@@ -48,7 +48,7 @@ movements:
         next: balthasar
   - name: balthasar
     persona: balthasar
-    instruction_template: |
+    instruction: |
       # MAGI System 継続
 
       ## 審議事項
@@ -82,7 +82,7 @@ movements:
         next: casper
   - name: casper
     persona: casper
-    instruction_template: |
+    instruction: |
       # MAGI System 最終審議
 
       ## 審議事項
diff --git a/builtins/ja/pieces/review-fix.yaml b/builtins/ja/pieces/review-fix.yaml
index 0d4d30d..f1c0249 100644
--- a/builtins/ja/pieces/review-fix.yaml
+++ b/builtins/ja/pieces/review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/review.yaml b/builtins/ja/pieces/review.yaml
index 9290200..b4f17ad 100644
--- a/builtins/ja/pieces/review.yaml
+++ b/builtins/ja/pieces/review.yaml
@@ -177,7 +177,7 @@ movements:
     rules:
       - condition: レビュー統合完了
         next: COMPLETE
-    instruction_template: |
+    instruction: |
       ## レビュー結果
       {previous_response}
 
diff --git a/builtins/ja/pieces/takt-default-review-fix.yaml b/builtins/ja/pieces/takt-default-review-fix.yaml
index 86d789c..d3b3388 100644
--- a/builtins/ja/pieces/takt-default-review-fix.yaml
+++ b/builtins/ja/pieces/takt-default-review-fix.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（進捗あり）
           next: reviewers
diff --git a/builtins/ja/pieces/takt-default-team-leader.yaml b/builtins/ja/pieces/takt-default-team-leader.yaml
index f4cb45c..cb0ae06 100644
--- a/builtins/ja/pieces/takt-default-team-leader.yaml
+++ b/builtins/ja/pieces/takt-default-team-leader.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-ai-fix
+      instruction: loop-monitor-ai-fix
       rules:
         - condition: 健全（進捗あり）
           next: ai_review
@@ -27,7 +27,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/takt-default.yaml b/builtins/ja/pieces/takt-default.yaml
index 4476b02..eebd5fb 100644
--- a/builtins/ja/pieces/takt-default.yaml
+++ b/builtins/ja/pieces/takt-default.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         ai_review と ai_fix のループが {cycle_count} 回繰り返されました。
 
         各サイクルのレポートを確認し、このループが健全（進捗がある）か、
@@ -39,7 +39,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: loop-monitor-reviewers-fix
+      instruction: loop-monitor-reviewers-fix
       rules:
         - condition: 健全（指摘数が減少、修正が反映されている）
           next: reviewers
diff --git a/builtins/ja/pieces/terraform.yaml b/builtins/ja/pieces/terraform.yaml
index 0f3c925..11e5445 100644
--- a/builtins/ja/pieces/terraform.yaml
+++ b/builtins/ja/pieces/terraform.yaml
@@ -265,7 +265,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         レビュー → 修正のサイクルが {cycle_count} 回繰り返されました。
         Report Directory 内のレビューレポート履歴を確認し、収束状況を判断してください。
 
@@ -282,7 +282,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         AI修正 → レビューのサイクルが {cycle_count} 回繰り返されました。
         Report Directory 内のレビューレポート履歴を確認し、収束状況を判断してください。
 
@@ -299,7 +299,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         監督修正 → レビューのサイクルが {cycle_count} 回繰り返されました。
         Report Directory 内のレビューレポート履歴を確認し、収束状況を判断してください。
 
diff --git a/builtins/ja/pieces/unit-test.yaml b/builtins/ja/pieces/unit-test.yaml
index c7fb8a4..3520a2b 100644
--- a/builtins/ja/pieces/unit-test.yaml
+++ b/builtins/ja/pieces/unit-test.yaml
@@ -15,7 +15,7 @@ loop_monitors:
     threshold: 3
     judge:
       persona: supervisor
-      instruction_template: |
+      instruction: |
         ai_review と ai_fix のループが {cycle_count} 回繰り返されました。
 
         各サイクルのレポートを確認し、このループが健全（進捗がある）か、
diff --git a/builtins/skill/references/yaml-schema.md b/builtins/skill/references/yaml-schema.md
index d9cadd1..61f9c40 100644
--- a/builtins/skill/references/yaml-schema.md
+++ b/builtins/skill/references/yaml-schema.md
@@ -47,22 +47,41 @@ movement 内では**キー名**で参照する（パスを直接書かない）
 - name: movement-name          # movement 名（必須、一意）
   persona: coder               # ペルソナキー（personas マップを参照、任意）
   policy: coding               # ポリシーキー（policies マップを参照、任意）
-  policy: [coding, testing]    # 複数指定も可（配列）
-  instruction: implement       # 指示テンプレートキー（instructions マップを参照、任意）
+  instruction: implement       # 指示（instructions マップのキー参照、またはインライン、任意）
   knowledge: architecture      # ナレッジキー（knowledge マップを参照、任意）
   edit: true                   # ファイル編集可否（必須）
   required_permission_mode: edit # 必要最小権限: edit / readonly / full（任意）
   session: refresh             # セッション管理（任意）
   pass_previous_response: true # 前の出力を渡すか（デフォルト: true）
   allowed_tools: [...]         # 許可ツール一覧（任意、参考情報）
-  instruction_template: |      # 指示テンプレート（参照解決またはインライン、任意）
-    指示内容...
   output_contracts: [...]      # 出力契約設定（任意）
   quality_gates: [...]         # 品質ゲート（AIへの指示、任意）
   rules: [...]                 # 遷移ルール（必須）
 ```
 
-**`instruction` vs `instruction_template`**: どちらも同じ参照解決ルート（セクションマップ → パス → 3-layer facet → インライン）を使う。`instruction_template` はインライン文字列もそのまま使える。通常はどちらか一方を使用する。
+複数ポリシー指定（配列）:
+
+```yaml
+- name: movement-name
+  policy: [coding, testing]
+```
+
+参照形式:
+
+```yaml
+- name: movement-name
+  instruction: implement
+```
+
+インライン形式:
+
+```yaml
+- name: movement-name
+  instruction: |
+    指示内容...
+```
+
+**`instruction`（正式） / `instruction_template`（deprecated）**: どちらも同じ参照解決ルート（セクションマップ → パス → 3-layer facet → インライン）を使う。`instruction_template` も互換のため受理されるが、新規定義では `instruction` を使う。
 
 ### Parallel Movement
 
@@ -185,7 +204,7 @@ quality_gates:
 
 ## テンプレート変数
 
-`instruction_template`（またはインストラクションファイル）内で使用可能な変数:
+`instruction`（またはインストラクションファイル）内で使用可能な変数:
 
 | 変数 | 説明 |
 |-----|------|
@@ -207,7 +226,7 @@ loop_monitors:
     threshold: 3                       # 発動閾値（サイクル回数）
     judge:
       persona: supervisor              # ペルソナキー参照
-      instruction_template: |          # 判定用指示
+      instruction: |                   # 判定用指示
         サイクルが {cycle_count} 回繰り返されました。
         健全性を判断してください。
       rules:
diff --git a/src/__tests__/aggregate-evaluator.test.ts b/src/__tests__/aggregate-evaluator.test.ts
index 9091d81..3509c59 100644
--- a/src/__tests__/aggregate-evaluator.test.ts
+++ b/src/__tests__/aggregate-evaluator.test.ts
@@ -35,7 +35,7 @@ function makeSubMovement(name: string, conditions: string[]): PieceMovement {
   return {
     name,
     personaDisplayName: name,
-    instructionTemplate: '',
+    instruction: '',
     passPreviousResponse: false,
     rules: conditions.map((c) => ({ condition: c })),
   };
@@ -48,7 +48,7 @@ function makeParentMovement(
   return {
     name: 'parent',
     personaDisplayName: 'parent',
-    instructionTemplate: '',
+    instruction: '',
     passPreviousResponse: false,
     parallel,
     rules,
@@ -264,7 +264,7 @@ describe('AggregateEvaluator', () => {
       const step: PieceMovement = {
         name: 'test-movement',
         personaDisplayName: 'tester',
-        instructionTemplate: '',
+        instruction: '',
         passPreviousResponse: false,
         rules: [
           {
diff --git a/src/__tests__/config.test.ts b/src/__tests__/config.test.ts
index bd4477d..9f618fa 100644
--- a/src/__tests__/config.test.ts
+++ b/src/__tests__/config.test.ts
@@ -77,7 +77,7 @@ describe('getBuiltinPiece', () => {
 
     const planMovement = piece!.movements.find((movement) => movement.name === 'plan');
     expect(planMovement).toBeDefined();
-    expect(planMovement!.instructionTemplate).not.toBe('plan');
+    expect(planMovement!.instruction).not.toBe('plan');
   });
 
   it('should return null for non-existent piece names', () => {
diff --git a/src/__tests__/engine-report.test.ts b/src/__tests__/engine-report.test.ts
index ad55b12..40e8aec 100644
--- a/src/__tests__/engine-report.test.ts
+++ b/src/__tests__/engine-report.test.ts
@@ -50,7 +50,7 @@ function createMovement(overrides: Partial<PieceMovement> = {}): PieceMovement {
     name: 'test-movement',
     persona: 'coder',
     personaDisplayName: 'Coder',
-    instructionTemplate: '',
+    instruction: '',
     passPreviousResponse: false,
     ...overrides,
   };
diff --git a/src/__tests__/engine-team-leader.test.ts b/src/__tests__/engine-team-leader.test.ts
index 0bbd5fb..6122fc1 100644
--- a/src/__tests__/engine-team-leader.test.ts
+++ b/src/__tests__/engine-team-leader.test.ts
@@ -32,7 +32,7 @@ function buildTeamLeaderConfig(): PieceConfig {
     maxMovements: 5,
     movements: [
       makeMovement('implement', {
-        instructionTemplate: 'Task: {task}',
+        instruction: 'Task: {task}',
         teamLeader: {
           persona: '../personas/team-leader.md',
           maxParts: 3,
diff --git a/src/__tests__/engine-test-helpers.ts b/src/__tests__/engine-test-helpers.ts
index 6198357..c40c55a 100644
--- a/src/__tests__/engine-test-helpers.ts
+++ b/src/__tests__/engine-test-helpers.ts
@@ -41,7 +41,7 @@ export function makeMovement(name: string, overrides: Partial<PieceMovement> = {
     name,
     persona: `../personas/${name}.md`,
     personaDisplayName: name,
-    instructionTemplate: `Run ${name}`,
+    instruction: `Run ${name}`,
     passPreviousResponse: true,
     ...overrides,
   };
diff --git a/src/__tests__/engine-worktree-report.test.ts b/src/__tests__/engine-worktree-report.test.ts
index 3cc2a8d..1f9c795 100644
--- a/src/__tests__/engine-worktree-report.test.ts
+++ b/src/__tests__/engine-worktree-report.test.ts
@@ -137,7 +137,7 @@ describe('PieceEngine: worktree reportDir resolution', () => {
       initialMovement: 'review',
       movements: [
         makeMovement('review', {
-          instructionTemplate: 'Write report to {report_dir}',
+          instruction: 'Write report to {report_dir}',
           outputContracts: [{ name: '00-review.md', format: '00-review', useJudge: true }],
           rules: [
             makeRule('approved', 'COMPLETE'),
diff --git a/src/__tests__/facet-resolution.test.ts b/src/__tests__/facet-resolution.test.ts
index 6e2acbe..6f98379 100644
--- a/src/__tests__/facet-resolution.test.ts
+++ b/src/__tests__/facet-resolution.test.ts
@@ -520,7 +520,6 @@ describe('normalizePieceConfig with layer resolution', () => {
           name: 'step1',
           persona: 'coder',
           instruction_template: 'implement',
-          instruction: '{task}',
         },
       ],
     };
@@ -528,7 +527,7 @@ describe('normalizePieceConfig with layer resolution', () => {
     const context: FacetResolutionContext = { projectDir, lang: 'ja' };
     const config = normalizePieceConfig(raw, pieceDir, context);
 
-    expect(config.movements[0]!.instructionTemplate).toBe('Mapped instruction template');
+    expect(config.movements[0]!.instruction).toBe('Mapped instruction template');
   });
 
   it('should resolve instruction_template by name via layer resolution', () => {
@@ -543,7 +542,6 @@ describe('normalizePieceConfig with layer resolution', () => {
           name: 'step1',
           persona: 'coder',
           instruction_template: 'implement',
-          instruction: '{task}',
         },
       ],
     };
@@ -551,7 +549,7 @@ describe('normalizePieceConfig with layer resolution', () => {
     const context: FacetResolutionContext = { projectDir, lang: 'ja' };
     const config = normalizePieceConfig(raw, pieceDir, context);
 
-    expect(config.movements[0]!.instructionTemplate).toBe('Project implement template');
+    expect(config.movements[0]!.instruction).toBe('Project implement template');
   });
 
   it('should keep inline instruction_template when no facet is found', () => {
@@ -564,7 +562,6 @@ Second line remains inline.`;
           name: 'step1',
           persona: 'coder',
           instruction_template: inlineTemplate,
-          instruction: '{task}',
         },
       ],
     };
@@ -572,7 +569,7 @@ Second line remains inline.`;
     const context: FacetResolutionContext = { projectDir, lang: 'ja' };
     const config = normalizePieceConfig(raw, pieceDir, context);
 
-    expect(config.movements[0]!.instructionTemplate).toBe(inlineTemplate);
+    expect(config.movements[0]!.instruction).toBe(inlineTemplate);
   });
 
   it('should resolve loop monitor judge instruction_template via layer resolution', () => {
@@ -612,6 +609,6 @@ Second line remains inline.`;
     const context: FacetResolutionContext = { projectDir, lang: 'ja' };
     const config = normalizePieceConfig(raw, pieceDir, context);
 
-    expect(config.loopMonitors?.[0]?.judge.instructionTemplate).toBe('Project judge template');
+    expect(config.loopMonitors?.[0]?.judge.instruction).toBe('Project judge template');
   });
 });
diff --git a/src/__tests__/instructionBuilder.test.ts b/src/__tests__/instructionBuilder.test.ts
index 2e036fd..cc9d770 100644
--- a/src/__tests__/instructionBuilder.test.ts
+++ b/src/__tests__/instructionBuilder.test.ts
@@ -32,7 +32,7 @@ function createMinimalStep(template: string): PieceMovement {
     name: 'test-step',
     persona: 'test-agent',
     personaDisplayName: 'Test Agent',
-    instructionTemplate: template,
+    instruction: template,
     passPreviousResponse: false,
   };
 }
diff --git a/src/__tests__/it-error-recovery.test.ts b/src/__tests__/it-error-recovery.test.ts
index 75199ba..a211cb8 100644
--- a/src/__tests__/it-error-recovery.test.ts
+++ b/src/__tests__/it-error-recovery.test.ts
@@ -63,7 +63,7 @@ function makeMovement(name: string, agentPath: string, rules: PieceRule[]): Piec
     persona: `./personas/${name}.md`,
     personaDisplayName: name,
     personaPath: agentPath,
-    instructionTemplate: '{task}',
+    instruction: '{task}',
     passPreviousResponse: true,
     rules,
   };
diff --git a/src/__tests__/it-instruction-builder.test.ts b/src/__tests__/it-instruction-builder.test.ts
index e972a3d..c5a1d87 100644
--- a/src/__tests__/it-instruction-builder.test.ts
+++ b/src/__tests__/it-instruction-builder.test.ts
@@ -40,7 +40,7 @@ function makeMovement(overrides: Partial<PieceMovement> = {}): PieceMovement {
     name: 'test-step',
     persona: 'test-agent',
     personaDisplayName: 'test-step',
-    instructionTemplate: 'Do the work.',
+    instruction: 'Do the work.',
     passPreviousResponse: false,
     rules: [
       makeRule('Done', 'COMPLETE'),
@@ -66,7 +66,7 @@ function makeContext(overrides: Partial<InstructionContext> = {}): InstructionCo
 
 describe('Instruction Builder IT: task auto-injection', () => {
   it('should auto-inject task as "User Request" section when template has no {task}', () => {
-    const step = makeMovement({ instructionTemplate: 'Do the work.' });
+    const step = makeMovement({ instruction: 'Do the work.' });
     const ctx = makeContext({ task: 'Build the login page' });
 
     const result = buildInstruction(step, ctx);
@@ -76,7 +76,7 @@ describe('Instruction Builder IT: task auto-injection', () => {
   });
 
   it('should NOT auto-inject task section when template contains {task}', () => {
-    const step = makeMovement({ instructionTemplate: 'Here is the task: {task}' });
+    const step = makeMovement({ instruction: 'Here is the task: {task}' });
     const ctx = makeContext({ task: 'Build the login page' });
 
     const result = buildInstruction(step, ctx);
@@ -93,7 +93,7 @@ describe('Instruction Builder IT: previous_response auto-injection', () => {
   it('should auto-inject previous response when passPreviousResponse is true', () => {
     const step = makeMovement({
       passPreviousResponse: true,
-      instructionTemplate: 'Continue the work.',
+      instruction: 'Continue the work.',
     });
     const previousOutput: AgentResponse = {
       persona: 'previous-agent',
@@ -112,7 +112,7 @@ describe('Instruction Builder IT: previous_response auto-injection', () => {
   it('should NOT inject previous response when passPreviousResponse is false', () => {
     const step = makeMovement({
       passPreviousResponse: false,
-      instructionTemplate: 'Do fresh work.',
+      instruction: 'Do fresh work.',
     });
     const previousOutput: AgentResponse = {
       persona: 'previous-agent',
@@ -131,7 +131,7 @@ describe('Instruction Builder IT: previous_response auto-injection', () => {
   it('should NOT auto-inject when template contains {previous_response}', () => {
     const step = makeMovement({
       passPreviousResponse: true,
-      instructionTemplate: '## Context\n{previous_response}\n\nDo work.',
+      instruction: '## Context\n{previous_response}\n\nDo work.',
     });
     const previousOutput: AgentResponse = {
       persona: 'prev', status: 'done', content: 'Prior work done.', timestamp: new Date(),
@@ -161,7 +161,7 @@ describe('Instruction Builder IT: user_inputs auto-injection', () => {
   });
 
   it('should NOT auto-inject when template contains {user_inputs}', () => {
-    const step = makeMovement({ instructionTemplate: 'Inputs: {user_inputs}' });
+    const step = makeMovement({ instruction: 'Inputs: {user_inputs}' });
     const ctx = makeContext({ userInputs: ['Input A'] });
 
     const result = buildInstruction(step, ctx);
@@ -175,7 +175,7 @@ describe('Instruction Builder IT: user_inputs auto-injection', () => {
 describe('Instruction Builder IT: iteration variables', () => {
   it('should replace {iteration}, {max_movements}, {movement_iteration} in template', () => {
     const step = makeMovement({
-      instructionTemplate: 'Iter: {iteration}/{max_movements}, movement iter: {movement_iteration}',
+      instruction: 'Iter: {iteration}/{max_movements}, movement iter: {movement_iteration}',
     });
     const ctx = makeContext({ iteration: 5, maxMovements: 30, movementIteration: 2 });
 
@@ -198,7 +198,7 @@ describe('Instruction Builder IT: iteration variables', () => {
 describe('Instruction Builder IT: report_dir expansion', () => {
   it('should replace {report_dir} in template', () => {
     const step = makeMovement({
-      instructionTemplate: 'Read the plan from {report_dir}/00-plan.md',
+      instruction: 'Read the plan from {report_dir}/00-plan.md',
     });
     const ctx = makeContext({ reportDir: '/tmp/test-project/.takt/runs/20250126-task/reports' });
 
@@ -209,7 +209,7 @@ describe('Instruction Builder IT: report_dir expansion', () => {
 
   it('should replace {report:filename} with full path', () => {
     const step = makeMovement({
-      instructionTemplate: 'Read {report:00-plan.md} for the plan.',
+      instruction: 'Read {report:00-plan.md} for the plan.',
     });
     const ctx = makeContext({ reportDir: '/tmp/reports' });
 
@@ -388,7 +388,7 @@ describe('Instruction Builder IT: template injection prevention', () => {
   it('should escape curly braces in previous response content', () => {
     const step = makeMovement({
       passPreviousResponse: true,
-      instructionTemplate: 'Continue.',
+      instruction: 'Continue.',
     });
     const ctx = makeContext({
       previousOutput: {
diff --git a/src/__tests__/it-notification-sound.test.ts b/src/__tests__/it-notification-sound.test.ts
index 459773e..df23647 100644
--- a/src/__tests__/it-notification-sound.test.ts
+++ b/src/__tests__/it-notification-sound.test.ts
@@ -224,7 +224,7 @@ function makeConfig(): PieceConfig {
         name: 'step1',
         persona: '../agents/coder.md',
         personaDisplayName: 'coder',
-        instructionTemplate: 'Do something',
+        instruction: 'Do something',
         passPreviousResponse: true,
         rules: [
           { condition: 'done', next: 'COMPLETE' },
diff --git a/src/__tests__/it-piece-execution.test.ts b/src/__tests__/it-piece-execution.test.ts
index 912fa8b..36e3b08 100644
--- a/src/__tests__/it-piece-execution.test.ts
+++ b/src/__tests__/it-piece-execution.test.ts
@@ -66,7 +66,7 @@ function makeMovement(name: string, agentPath: string, rules: PieceRule[]): Piec
     persona: `./personas/${name}.md`,
     personaDisplayName: name,
     personaPath: agentPath,
-    instructionTemplate: '{task}',
+    instruction: '{task}',
     passPreviousResponse: true,
     rules,
   };
diff --git a/src/__tests__/it-piece-loader.test.ts b/src/__tests__/it-piece-loader.test.ts
index a0d1bb0..1c64aaf 100644
--- a/src/__tests__/it-piece-loader.test.ts
+++ b/src/__tests__/it-piece-loader.test.ts
@@ -92,8 +92,8 @@ describe('Piece Loader IT: builtin piece loading', () => {
 
     expect(planMovement).toBeDefined();
     expect(implementMovement).toBeDefined();
-    expect(planMovement!.instructionTemplate).toContain('missing E2E tests');
-    expect(implementMovement!.instructionTemplate).toContain('npm run test:e2e:mock');
+    expect(planMovement!.instruction).toContain('missing E2E tests');
+    expect(implementMovement!.instruction).toContain('npm run test:e2e:mock');
   });
 
   it('should load e2e-test as a builtin piece in ja locale', () => {
@@ -110,8 +110,8 @@ describe('Piece Loader IT: builtin piece loading', () => {
 
     expect(planMovement).toBeDefined();
     expect(implementMovement).toBeDefined();
-    expect(planMovement!.instructionTemplate).toContain('E2Eテスト');
-    expect(implementMovement!.instructionTemplate).toContain('npm run test:e2e:mock');
+    expect(planMovement!.instruction).toContain('E2Eテスト');
+    expect(implementMovement!.instruction).toContain('npm run test:e2e:mock');
   });
 });
 
@@ -156,6 +156,49 @@ movements:
     expect(config!.movements.length).toBe(1);
     expect(config!.movements[0]!.name).toBe('start');
   });
+
+  it('should propagate canonical instruction field through loader for movement and loop monitor judge', () => {
+    // Given: project-local piece that uses instruction on both movement and loop monitor judge
+    const piecesDir = join(testDir, '.takt', 'pieces');
+    mkdirSync(piecesDir, { recursive: true });
+
+    writeFileSync(join(piecesDir, 'instruction-canonical.yaml'), `
+name: instruction-canonical
+max_movements: 8
+initial_movement: step1
+
+movements:
+  - name: step1
+    instruction: "Step 1 instruction"
+    rules:
+      - condition: next
+        next: step2
+  - name: step2
+    instruction: "Step 2 instruction"
+    rules:
+      - condition: done
+        next: COMPLETE
+
+loop_monitors:
+  - cycle: [step1, step2]
+    threshold: 2
+    judge:
+      instruction: "Judge instruction"
+      rules:
+        - condition: continue
+          next: step2
+`);
+
+    // When: loading the piece through the integration entry point
+    const config = loadPiece('instruction-canonical', testDir);
+
+    // Then: canonical instruction is available on normalized movement/judge models
+    expect(config).not.toBeNull();
+    const step1 = config!.movements[0] as unknown as Record<string, unknown>;
+    const judge = config!.loopMonitors?.[0]?.judge as unknown as Record<string, unknown>;
+    expect(step1.instruction).toBe('Step 1 instruction');
+    expect(judge.instruction).toBe('Judge instruction');
+  });
 });
 
 describe('Piece Loader IT: agent path resolution', () => {
diff --git a/src/__tests__/it-rule-evaluation.test.ts b/src/__tests__/it-rule-evaluation.test.ts
index 57ec9cc..a0da237 100644
--- a/src/__tests__/it-rule-evaluation.test.ts
+++ b/src/__tests__/it-rule-evaluation.test.ts
@@ -50,7 +50,7 @@ function makeMovement(
     name,
     persona: 'test-agent',
     personaDisplayName: name,
-    instructionTemplate: '{task}',
+    instruction: '{task}',
     passPreviousResponse: true,
     rules,
     parallel,
@@ -401,7 +401,7 @@ describe('Rule Evaluation IT: movements without rules', () => {
       name: 'step',
       persona: 'agent',
       personaDisplayName: 'step',
-      instructionTemplate: '{task}',
+      instruction: '{task}',
       passPreviousResponse: true,
     };
 
diff --git a/src/__tests__/it-sigint-interrupt.test.ts b/src/__tests__/it-sigint-interrupt.test.ts
index 114b733..c8b75d9 100644
--- a/src/__tests__/it-sigint-interrupt.test.ts
+++ b/src/__tests__/it-sigint-interrupt.test.ts
@@ -224,7 +224,7 @@ describe('executePiece: SIGINT handler integration', () => {
           name: 'step1',
           persona: '../agents/coder.md',
           personaDisplayName: 'coder',
-          instructionTemplate: 'Do something',
+          instruction: 'Do something',
           passPreviousResponse: true,
           rules: [
             { condition: 'done', next: 'COMPLETE' },
diff --git a/src/__tests__/it-three-phase-execution.test.ts b/src/__tests__/it-three-phase-execution.test.ts
index 82338a4..5ae29da 100644
--- a/src/__tests__/it-three-phase-execution.test.ts
+++ b/src/__tests__/it-three-phase-execution.test.ts
@@ -93,7 +93,7 @@ function makeMovement(
     persona: './agents/agent.md',
     personaDisplayName: name,
     personaPath: agentPath,
-    instructionTemplate: '{task}',
+    instruction: '{task}',
     passPreviousResponse: true,
     rules,
     outputContracts: options.outputContracts,
diff --git a/src/__tests__/knowledge.test.ts b/src/__tests__/knowledge.test.ts
index baa8839..43cf15e 100644
--- a/src/__tests__/knowledge.test.ts
+++ b/src/__tests__/knowledge.test.ts
@@ -317,11 +317,11 @@ describe('normalizePieceConfig knowledge resolution', () => {
 
 // --- Test helpers for InstructionBuilder ---
 
-function createMinimalStep(instructionTemplate: string): PieceMovement {
+function createMinimalStep(instruction: string): PieceMovement {
   return {
     name: 'test-step',
     personaDisplayName: 'coder',
-    instructionTemplate,
+    instruction,
     passPreviousResponse: false,
   };
 }
diff --git a/src/__tests__/options-builder.test.ts b/src/__tests__/options-builder.test.ts
index 478b47d..d940b79 100644
--- a/src/__tests__/options-builder.test.ts
+++ b/src/__tests__/options-builder.test.ts
@@ -7,7 +7,7 @@ function createMovement(overrides: Partial<PieceMovement> = {}): PieceMovement {
   return {
     name: 'reviewers',
     personaDisplayName: 'Reviewers',
-    instructionTemplate: 'review',
+    instruction: 'review',
     passPreviousResponse: false,
     ...overrides,
   };
diff --git a/src/__tests__/parallel-and-loader.test.ts b/src/__tests__/parallel-and-loader.test.ts
index 4c484e7..3d72341 100644
--- a/src/__tests__/parallel-and-loader.test.ts
+++ b/src/__tests__/parallel-and-loader.test.ts
@@ -8,7 +8,12 @@
  */
 
 import { describe, it, expect } from 'vitest';
-import { PieceConfigRawSchema, ParallelSubMovementRawSchema, PieceMovementRawSchema } from '../core/models/index.js';
+import {
+  PieceConfigRawSchema,
+  ParallelSubMovementRawSchema,
+  PieceMovementRawSchema,
+  LoopMonitorJudgeSchema,
+} from '../core/models/index.js';
 
 describe('ParallelSubMovementRawSchema', () => {
   it('should validate a valid parallel sub-movement', () => {
@@ -32,6 +37,39 @@ describe('ParallelSubMovementRawSchema', () => {
     expect(result.success).toBe(true);
   });
 
+  it('should accept a sub-movement with instruction field', () => {
+    // Given: a parallel sub-movement that uses the new canonical field
+    const raw = {
+      name: 'no-agent-step',
+      instruction: 'Do something',
+    };
+
+    // When: validating the sub-movement schema
+    const result = ParallelSubMovementRawSchema.safeParse(raw);
+
+    // Then: it is accepted
+    expect(result.success).toBe(true);
+  });
+
+  it('should accept a sub-movement when instruction and instruction_template are both provided', () => {
+    // Given: both canonical and deprecated fields are present during migration
+    const raw = {
+      name: 'dual-field-sub-step',
+      instruction: 'Canonical instruction',
+      instruction_template: 'Legacy instruction',
+    };
+
+    // When: validating the sub-movement schema
+    const result = ParallelSubMovementRawSchema.safeParse(raw);
+
+    // Then: schema keeps backward compatibility and accepts both fields
+    expect(result.success).toBe(true);
+    if (result.success) {
+      expect((result.data as unknown as Record<string, unknown>).instruction).toBe('Canonical instruction');
+      expect((result.data as unknown as Record<string, unknown>).instruction_template).toBe('Legacy instruction');
+    }
+  });
+
   it('should accept optional fields', () => {
     const raw = {
       name: 'full-sub-step',
@@ -144,6 +182,39 @@ describe('PieceMovementRawSchema with parallel', () => {
     expect(result.success).toBe(true);
   });
 
+  it('should accept a movement with instruction only', () => {
+    // Given: a movement that only uses instruction
+    const raw = {
+      name: 'orphan-step',
+      instruction: 'Do something',
+    };
+
+    // When: validating the movement schema
+    const result = PieceMovementRawSchema.safeParse(raw);
+
+    // Then: it is accepted
+    expect(result.success).toBe(true);
+  });
+
+  it('should accept a movement when instruction and instruction_template are both provided', () => {
+    // Given: movement includes both canonical and deprecated instruction fields
+    const raw = {
+      name: 'orphan-step',
+      instruction: 'Canonical movement instruction',
+      instruction_template: 'Legacy movement instruction',
+    };
+
+    // When: validating the movement schema
+    const result = PieceMovementRawSchema.safeParse(raw);
+
+    // Then: schema accepts both fields for deprecation window
+    expect(result.success).toBe(true);
+    if (result.success) {
+      expect((result.data as unknown as Record<string, unknown>).instruction).toBe('Canonical movement instruction');
+      expect((result.data as unknown as Record<string, unknown>).instruction_template).toBe('Legacy movement instruction');
+    }
+  });
+
   it('should accept a movement with persona (no parallel)', () => {
     const raw = {
       name: 'normal-step',
@@ -182,6 +253,46 @@ describe('PieceMovementRawSchema with parallel', () => {
   });
 });
 
+describe('LoopMonitorJudgeSchema', () => {
+  it('should accept judge configuration with instruction field', () => {
+    // Given: a loop monitor judge with canonical instruction
+    const raw = {
+      persona: 'reviewer',
+      instruction: 'Judge loop health',
+      rules: [{ condition: 'continue', next: 'ai_fix' }],
+    };
+
+    // When: validating judge schema
+    const result = LoopMonitorJudgeSchema.safeParse(raw);
+
+    // Then: it is accepted
+    expect(result.success).toBe(true);
+    if (result.success) {
+      expect((result.data as unknown as Record<string, unknown>).instruction).toBe('Judge loop health');
+    }
+  });
+
+  it('should accept judge configuration during deprecation window when both fields exist', () => {
+    // Given: judge config with both new and deprecated fields
+    const raw = {
+      persona: 'reviewer',
+      instruction: 'Judge loop health',
+      instruction_template: 'legacy judge instruction',
+      rules: [{ condition: 'continue', next: 'ai_fix' }],
+    };
+
+    // When: validating judge schema
+    const result = LoopMonitorJudgeSchema.safeParse(raw);
+
+    // Then: it is accepted for backward compatibility
+    expect(result.success).toBe(true);
+    if (result.success) {
+      expect((result.data as unknown as Record<string, unknown>).instruction).toBe('Judge loop health');
+      expect((result.data as unknown as Record<string, unknown>).instruction_template).toBe('legacy judge instruction');
+    }
+  });
+});
+
 describe('PieceConfigRawSchema with parallel movements', () => {
   it('should validate a piece with parallel movement', () => {
     const raw = {
diff --git a/src/__tests__/phase-runner-report-history.test.ts b/src/__tests__/phase-runner-report-history.test.ts
index 377af91..92e0fba 100644
--- a/src/__tests__/phase-runner-report-history.test.ts
+++ b/src/__tests__/phase-runner-report-history.test.ts
@@ -17,7 +17,7 @@ function createStep(fileName: string): PieceMovement {
   return {
     name: 'reviewers',
     personaDisplayName: 'Reviewers',
-    instructionTemplate: 'review',
+    instruction: 'review',
     passPreviousResponse: false,
     outputContracts: [{ name: fileName }],
   };
diff --git a/src/__tests__/pieceExecution-ask-user-question.test.ts b/src/__tests__/pieceExecution-ask-user-question.test.ts
index 0ef0ed8..2f68cb7 100644
--- a/src/__tests__/pieceExecution-ask-user-question.test.ts
+++ b/src/__tests__/pieceExecution-ask-user-question.test.ts
@@ -49,7 +49,7 @@ const { MockPieceEngine } = vi.hoisted(() => {
         return { status: 'aborted', iteration: 1 };
       }
       if (firstStep) {
-        this.emit('movement:start', firstStep, 1, firstStep.instructionTemplate, { provider: undefined, model: undefined });
+        this.emit('movement:start', firstStep, 1, firstStep.instruction, { provider: undefined, model: undefined });
       }
       this.emit('piece:complete', { status: 'completed', iteration: 1 });
       return { status: 'completed', iteration: 1 };
@@ -171,7 +171,7 @@ function makeConfig(): PieceConfig {
         name: 'implement',
         persona: '../agents/coder.md',
         personaDisplayName: 'coder',
-        instructionTemplate: 'Implement task',
+        instruction: 'Implement task',
         passPreviousResponse: true,
         rules: [{ condition: 'done', next: 'COMPLETE' }],
       },
diff --git a/src/__tests__/pieceExecution-debug-prompts.test.ts b/src/__tests__/pieceExecution-debug-prompts.test.ts
index 6e7bf12..95a5384 100644
--- a/src/__tests__/pieceExecution-debug-prompts.test.ts
+++ b/src/__tests__/pieceExecution-debug-prompts.test.ts
@@ -236,7 +236,7 @@ describe('executePiece debug prompts logging', () => {
           name: 'implement',
           persona: '../agents/coder.md',
           personaDisplayName: 'coder',
-          instructionTemplate: 'Implement task',
+          instruction: 'Implement task',
           passPreviousResponse: true,
           rules: [{ condition: 'done', next: 'COMPLETE' }],
         },
diff --git a/src/__tests__/pieceExecution-session-loading.test.ts b/src/__tests__/pieceExecution-session-loading.test.ts
index 98dd9a8..e93de57 100644
--- a/src/__tests__/pieceExecution-session-loading.test.ts
+++ b/src/__tests__/pieceExecution-session-loading.test.ts
@@ -77,7 +77,7 @@ const {
       const firstStep = this.config.movements[0];
       if (firstStep) {
         const providerInfo = resolveProviderInfo(firstStep, this.receivedOptions);
-        this.emit('movement:start', firstStep, 1, firstStep.instructionTemplate, providerInfo);
+        this.emit('movement:start', firstStep, 1, firstStep.instruction, providerInfo);
         this.emit('movement:complete', firstStep, {
           persona: firstStep.personaDisplayName,
           status: 'done',
@@ -85,7 +85,7 @@ const {
           timestamp: new Date('2026-03-04T00:00:00.000Z'),
           sessionId: 'movement-session',
           providerUsage: mockMovementResponse.providerUsage,
-        }, firstStep.instructionTemplate);
+        }, firstStep.instruction);
       }
       this.emit('piece:complete', { status: 'completed', iteration: 1 });
       return { status: 'completed', iteration: 1 };
@@ -230,7 +230,7 @@ function makeConfig(): PieceConfig {
         name: 'implement',
         persona: '../agents/coder.md',
         personaDisplayName: 'coder',
-        instructionTemplate: 'Implement task',
+        instruction: 'Implement task',
         passPreviousResponse: true,
         rules: [{ condition: 'done', next: 'COMPLETE' }],
       },
diff --git a/src/__tests__/policy-persona.test.ts b/src/__tests__/policy-persona.test.ts
index f841b2c..52c684d 100644
--- a/src/__tests__/policy-persona.test.ts
+++ b/src/__tests__/policy-persona.test.ts
@@ -9,7 +9,7 @@
  * - File-based policy content loading via resolveContentPath
  */
 
-import { describe, it, expect, beforeEach, afterEach } from 'vitest';
+import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest';
 import { mkdtempSync, mkdirSync, writeFileSync, rmSync } from 'node:fs';
 import { join } from 'node:path';
 import { tmpdir } from 'node:os';
@@ -371,7 +371,7 @@ describe('InstructionBuilder policy injection', () => {
     const step = {
       name: 'test-step',
       personaDisplayName: 'coder',
-      instructionTemplate: 'Do the thing.',
+      instruction: 'Do the thing.',
       passPreviousResponse: false,
       policyContents: ['# Coding Policy\n\nWrite clean code.'],
     };
@@ -390,7 +390,7 @@ describe('InstructionBuilder policy injection', () => {
     const step = {
       name: 'test-step',
       personaDisplayName: 'coder',
-      instructionTemplate: 'Do the thing.',
+      instruction: 'Do the thing.',
       passPreviousResponse: false,
       policyContents: ['# Coding Policy\n\nWrite clean code.'],
     };
@@ -408,7 +408,7 @@ describe('InstructionBuilder policy injection', () => {
     const step = {
       name: 'test-step',
       personaDisplayName: 'coder',
-      instructionTemplate: 'Do the thing.',
+      instruction: 'Do the thing.',
       passPreviousResponse: false,
     };
 
@@ -423,7 +423,7 @@ describe('InstructionBuilder policy injection', () => {
     const step = {
       name: 'test-step',
       personaDisplayName: 'coder',
-      instructionTemplate: 'Do the thing.',
+      instruction: 'Do the thing.',
       passPreviousResponse: false,
       policyContents: ['Policy A content.', 'Policy B content.'],
     };
@@ -441,7 +441,7 @@ describe('InstructionBuilder policy injection', () => {
     const step = {
       name: 'test-step',
       personaDisplayName: 'coder',
-      instructionTemplate: 'Do the thing.',
+      instruction: 'Do the thing.',
       passPreviousResponse: false,
       policyContents: ['Step policy.'],
     };
@@ -545,7 +545,22 @@ describe('section reference resolution', () => {
     };
 
     const config = normalizePieceConfig(raw, testDir);
-    expect(config.movements[0]!.instructionTemplate).toBe('Implement the feature.');
+    expect(config.movements[0]!.instruction).toBe('Implement the feature.');
+  });
+
+  it('should expose normalized movement instruction on instruction field', () => {
+    const raw = {
+      name: 'test-piece',
+      movements: [{
+        name: 'impl',
+        persona: 'coder',
+        instruction: 'Canonical movement instruction',
+      }],
+    };
+
+    const config = normalizePieceConfig(raw, testDir);
+    const movement = config.movements[0] as unknown as Record<string, unknown>;
+    expect(movement.instruction).toBe('Canonical movement instruction');
   });
 
   it('should resolve output contract from report_formats section by name', () => {
@@ -586,7 +601,7 @@ describe('section reference resolution', () => {
     expect(config.movements[0]!.persona).toBe('nonexistent');
   });
 
-  it('should prefer instruction_template over instruction section reference', () => {
+  it('should prefer instruction over instruction_template when both are provided', () => {
     const raw = {
       name: 'test-piece',
       instructions: { implement: './instructions/implement.md' },
@@ -599,7 +614,144 @@ describe('section reference resolution', () => {
     };
 
     const config = normalizePieceConfig(raw, testDir);
-    expect(config.movements[0]!.instructionTemplate).toBe('Inline template takes priority.');
+    expect(config.movements[0]!.instruction).toBe('Implement the feature.');
+  });
+
+  it('should emit deprecation warning when movement uses instruction_template', () => {
+    // Given: deprecated instruction_template is used on a movement
+    const warnSpy = vi.spyOn(console, 'warn').mockImplementation(() => {});
+    try {
+      const raw = {
+        name: 'test-piece',
+        movements: [{
+          name: 'impl',
+          persona: 'coder',
+          instruction_template: 'Legacy movement instruction',
+        }],
+      };
+
+      // When: normalizing piece config
+      normalizePieceConfig(raw, testDir);
+
+      // Then: deprecation warning is emitted
+      expect(warnSpy).toHaveBeenCalledWith(expect.stringContaining('instruction_template'));
+    } finally {
+      warnSpy.mockRestore();
+    }
+  });
+
+  it('should emit deprecation warning when loop monitor judge uses instruction_template', () => {
+    // Given: deprecated instruction_template is used on loop monitor judge
+    const warnSpy = vi.spyOn(console, 'warn').mockImplementation(() => {});
+    try {
+      const raw = {
+        name: 'test-piece',
+        movements: [
+          {
+            name: 'step1',
+            persona: 'coder',
+            instruction: '{task}',
+            rules: [{ condition: 'next', next: 'step2' }],
+          },
+          {
+            name: 'step2',
+            persona: 'coder',
+            instruction: '{task}',
+            rules: [{ condition: 'done', next: 'COMPLETE' }],
+          },
+        ],
+        loop_monitors: [
+          {
+            cycle: ['step1', 'step2'],
+            threshold: 2,
+            judge: {
+              persona: 'coder',
+              instruction_template: 'Legacy judge instruction',
+              rules: [{ condition: 'continue', next: 'step2' }],
+            },
+          },
+        ],
+      };
+
+      // When: normalizing piece config
+      normalizePieceConfig(raw, testDir);
+
+      // Then: deprecation warning is emitted
+      expect(warnSpy).toHaveBeenCalledWith(expect.stringContaining('instruction_template'));
+    } finally {
+      warnSpy.mockRestore();
+    }
+  });
+
+  it('should prefer loop monitor judge instruction over instruction_template when both are provided', () => {
+    const raw = {
+      name: 'test-piece',
+      instructions: { judge_template: './instructions/implement.md' },
+      movements: [
+        {
+          name: 'step1',
+          persona: 'coder',
+          instruction: '{task}',
+          rules: [{ condition: 'next', next: 'step2' }],
+        },
+        {
+          name: 'step2',
+          persona: 'coder',
+          instruction: '{task}',
+          rules: [{ condition: 'done', next: 'COMPLETE' }],
+        },
+      ],
+      loop_monitors: [
+        {
+          cycle: ['step1', 'step2'],
+          threshold: 2,
+          judge: {
+            persona: 'coder',
+            instruction: 'judge_template',
+            instruction_template: 'Legacy judge template',
+            rules: [{ condition: 'continue', next: 'step2' }],
+          },
+        },
+      ],
+    };
+
+    const config = normalizePieceConfig(raw, testDir);
+    expect(config.loopMonitors?.[0]?.judge.instruction).toBe('Implement the feature.');
+  });
+
+  it('should expose normalized loop monitor judge instruction on instruction field', () => {
+    const raw = {
+      name: 'test-piece',
+      movements: [
+        {
+          name: 'step1',
+          persona: 'coder',
+          instruction: '{task}',
+          rules: [{ condition: 'next', next: 'step2' }],
+        },
+        {
+          name: 'step2',
+          persona: 'coder',
+          instruction: '{task}',
+          rules: [{ condition: 'done', next: 'COMPLETE' }],
+        },
+      ],
+      loop_monitors: [
+        {
+          cycle: ['step1', 'step2'],
+          threshold: 2,
+          judge: {
+            persona: 'coder',
+            instruction: 'Canonical judge instruction',
+            rules: [{ condition: 'continue', next: 'step2' }],
+          },
+        },
+      ],
+    };
+
+    const config = normalizePieceConfig(raw, testDir);
+    const judge = config.loopMonitors?.[0]?.judge as unknown as Record<string, unknown>;
+    expect(judge.instruction).toBe('Canonical judge instruction');
   });
 
   it('should store resolved sections on PieceConfig', () => {
@@ -653,7 +805,7 @@ describe('section reference resolution', () => {
     const parallel = config.movements[0]!.parallel!;
     expect(parallel[0]!.persona).toBe('./personas/coder.md');
     expect(parallel[0]!.policyContents).toEqual(['# Coding Policy\nWrite clean code.']);
-    expect(parallel[0]!.instructionTemplate).toBe('Implement the feature.');
+    expect(parallel[0]!.instruction).toBe('Implement the feature.');
     expect(parallel[1]!.policyContents).toEqual([
       '# Coding Policy\nWrite clean code.',
       '# Testing Policy\nTest everything.',
diff --git a/src/__tests__/report-phase-retry.test.ts b/src/__tests__/report-phase-retry.test.ts
index 74934a5..a46b7f7 100644
--- a/src/__tests__/report-phase-retry.test.ts
+++ b/src/__tests__/report-phase-retry.test.ts
@@ -17,7 +17,7 @@ function createStep(fileName: string): PieceMovement {
     name: 'implement',
     persona: 'coder',
     personaDisplayName: 'Coder',
-    instructionTemplate: 'Implement task',
+    instruction: 'Implement task',
     passPreviousResponse: false,
     outputContracts: [{ name: fileName }],
   };
diff --git a/src/__tests__/session-key.test.ts b/src/__tests__/session-key.test.ts
index be4ccc1..13e63ab 100644
--- a/src/__tests__/session-key.test.ts
+++ b/src/__tests__/session-key.test.ts
@@ -11,7 +11,7 @@ function createMovement(overrides: Partial<PieceMovement> = {}): PieceMovement {
     name: 'test-movement',
     personaDisplayName: 'test',
     edit: false,
-    instructionTemplate: '',
+    instruction: '',
     passPreviousResponse: true,
     ...overrides,
   };
diff --git a/src/__tests__/status-judgment-phase.test.ts b/src/__tests__/status-judgment-phase.test.ts
index 6fb91ba..6648206 100644
--- a/src/__tests__/status-judgment-phase.test.ts
+++ b/src/__tests__/status-judgment-phase.test.ts
@@ -43,7 +43,7 @@ describe('runStatusJudgmentPhase', () => {
       name: 'review',
       persona: 'reviewer',
       personaDisplayName: 'reviewer',
-      instructionTemplate: 'Review',
+      instruction: 'Review',
       passPreviousResponse: true,
       rules: [
         { condition: 'needs_fix', next: 'fix' },
@@ -103,7 +103,7 @@ describe('runStatusJudgmentPhase', () => {
       name: 'review',
       persona: 'reviewer',
       personaDisplayName: 'reviewer',
-      instructionTemplate: 'Review',
+      instruction: 'Review',
       passPreviousResponse: true,
       rules: [
         { condition: 'needs_fix', next: 'fix' },
diff --git a/src/__tests__/team-leader-common.test.ts b/src/__tests__/team-leader-common.test.ts
index ab196c5..106714b 100644
--- a/src/__tests__/team-leader-common.test.ts
+++ b/src/__tests__/team-leader-common.test.ts
@@ -9,7 +9,7 @@ describe('createPartMovement', () => {
       name: 'implement',
       persona: 'coder',
       personaDisplayName: 'Coder',
-      instructionTemplate: 'do work',
+      instruction: 'do work',
       passPreviousResponse: false,
       providerOptions: {
         claude: {
diff --git a/src/__tests__/test-helpers.ts b/src/__tests__/test-helpers.ts
index e6118df..1392fda 100644
--- a/src/__tests__/test-helpers.ts
+++ b/src/__tests__/test-helpers.ts
@@ -16,7 +16,7 @@ export function makeMovement(overrides: Partial<PieceMovement> = {}): PieceMovem
   return {
     name: 'test-movement',
     personaDisplayName: 'tester',
-    instructionTemplate: '',
+    instruction: '',
     passPreviousResponse: false,
     ...overrides,
   };
diff --git a/src/__tests__/transitions.test.ts b/src/__tests__/transitions.test.ts
index 4bf3244..a856367 100644
--- a/src/__tests__/transitions.test.ts
+++ b/src/__tests__/transitions.test.ts
@@ -12,7 +12,7 @@ function createMovementWithRules(rules: { condition: string; next: string }[]):
     name: 'test-step',
     persona: 'test-agent',
     personaDisplayName: 'Test Agent',
-    instructionTemplate: '{task}',
+    instruction: '{task}',
     passPreviousResponse: false,
     rules: rules.map((r) => ({
       condition: r.condition,
@@ -47,7 +47,7 @@ describe('determineNextMovementByRules', () => {
       name: 'test-step',
       persona: 'test-agent',
       personaDisplayName: 'Test Agent',
-      instructionTemplate: '{task}',
+      instruction: '{task}',
       passPreviousResponse: false,
     };
 
@@ -68,7 +68,7 @@ describe('determineNextMovementByRules', () => {
       name: 'sub-step',
       persona: 'test-agent',
       personaDisplayName: 'Test Agent',
-      instructionTemplate: '{task}',
+      instruction: '{task}',
       passPreviousResponse: false,
       rules: [
         { condition: 'approved' },
diff --git a/src/__tests__/worktree-exceeded-requeue.test.ts b/src/__tests__/worktree-exceeded-requeue.test.ts
index 4fa4571..6871eb5 100644
--- a/src/__tests__/worktree-exceeded-requeue.test.ts
+++ b/src/__tests__/worktree-exceeded-requeue.test.ts
@@ -126,7 +126,7 @@ function buildTestPieceConfig(): PieceConfig {
         name: 'plan',
         persona: '../personas/plan.md',
         personaDisplayName: 'plan',
-        instructionTemplate: 'Run plan',
+        instruction: 'Run plan',
         passPreviousResponse: true,
         rules: [],
       },
diff --git a/src/__tests__/yaml-schema-reference.test.ts b/src/__tests__/yaml-schema-reference.test.ts
new file mode 100644
index 0000000..8fd14b1
--- /dev/null
+++ b/src/__tests__/yaml-schema-reference.test.ts
@@ -0,0 +1,47 @@
+import { describe, it, expect } from 'vitest';
+import { readFileSync } from 'node:fs';
+import { join } from 'node:path';
+
+describe('yaml-schema reference', () => {
+  it('normal movement examples should not duplicate instruction keys in one movement block', () => {
+    const schemaPath = join(process.cwd(), 'builtins', 'skill', 'references', 'yaml-schema.md');
+    const schemaText = readFileSync(schemaPath, 'utf-8');
+    const normalSectionMatch = schemaText.match(/### 通常 Movement([\s\S]*?)### Parallel Movement/);
+
+    expect(normalSectionMatch).not.toBeNull();
+    const normalSection = normalSectionMatch![1];
+    const yamlBlocks = [...normalSection.matchAll(/```yaml\n([\s\S]*?)```/g)].map((m) => m[1]);
+
+    expect(yamlBlocks.length).toBeGreaterThan(0);
+
+    const hasDuplicatedInstruction = yamlBlocks.some((block) => {
+      const instructionKeys = block
+        .split('\n')
+        .filter((line) => /^ {2}instruction:\s/.test(line));
+      return instructionKeys.length > 1;
+    });
+
+    expect(hasDuplicatedInstruction).toBe(false);
+  });
+
+  it('normal movement examples should not duplicate policy keys in one movement block', () => {
+    const schemaPath = join(process.cwd(), 'builtins', 'skill', 'references', 'yaml-schema.md');
+    const schemaText = readFileSync(schemaPath, 'utf-8');
+    const normalSectionMatch = schemaText.match(/### 通常 Movement([\s\S]*?)### Parallel Movement/);
+
+    expect(normalSectionMatch).not.toBeNull();
+    const normalSection = normalSectionMatch![1];
+    const yamlBlocks = [...normalSection.matchAll(/```yaml\n([\s\S]*?)```/g)].map((m) => m[1]);
+
+    expect(yamlBlocks.length).toBeGreaterThan(0);
+
+    const hasDuplicatedPolicy = yamlBlocks.some((block) => {
+      const policyKeys = block
+        .split('\n')
+        .filter((line) => /^ {2}policy:\s/.test(line));
+      return policyKeys.length > 1;
+    });
+
+    expect(hasDuplicatedPolicy).toBe(false);
+  });
+});
diff --git a/src/core/models/piece-types.ts b/src/core/models/piece-types.ts
index 98680c4..ba86b47 100644
--- a/src/core/models/piece-types.ts
+++ b/src/core/models/piece-types.ts
@@ -42,7 +42,7 @@ export interface OutputContractItem {
   format: string;
   /** Whether this report is used as input for status judgment phase (default: true) */
   useJudge?: boolean;
-  /** Instruction prepended before instruction_template (e.g., output destination) */
+  /** Instruction prepended before movement instruction (e.g., output destination) */
   order?: string;
 }
 
@@ -143,7 +143,7 @@ export interface PieceMovement {
   providerOptions?: MovementProviderOptions;
   /** Whether this movement is allowed to edit project files (true=allowed, false=prohibited, undefined=no prompt) */
   edit?: boolean;
-  instructionTemplate: string;
+  instruction: string;
   /** Rules for movement routing */
   rules?: PieceRule[];
   /** Output contracts for this movement (report definitions) */
@@ -219,8 +219,8 @@ export interface LoopMonitorJudge {
   persona?: string;
   /** Resolved absolute path to persona prompt file (set by loader) */
   personaPath?: string;
-  /** Custom instruction template for the judge (uses default if omitted) */
-  instructionTemplate?: string;
+  /** Custom instruction for the judge (uses default if omitted) */
+  instruction?: string;
   /** Rules for the judge's decision */
   rules: LoopMonitorRule[];
 }
diff --git a/src/core/models/schemas.ts b/src/core/models/schemas.ts
index 22d4802..842ab9c 100644
--- a/src/core/models/schemas.ts
+++ b/src/core/models/schemas.ts
@@ -162,11 +162,11 @@ export const PieceProviderOptionsSchema = z.object({
 export const OutputContractItemSchema = z.object({
   /** Report file name */
   name: z.string().min(1),
-  /** Instruction appended after instruction_template (e.g., output format) */
+  /** Instruction appended after movement instruction (e.g., output format) */
   format: z.string().min(1),
   /** Whether this report is used as input for status judgment phase */
   use_judge: z.boolean().optional().default(true),
-  /** Instruction prepended before instruction_template (e.g., output destination) */
+  /** Instruction prepended before movement instruction (e.g., output destination) */
   order: z.string().optional(),
 });
 
@@ -376,7 +376,9 @@ export const LoopMonitorRuleSchema = z.object({
 export const LoopMonitorJudgeSchema = z.object({
   /** Persona reference — key name from piece-level personas map, or file path */
   persona: z.string().optional(),
-  /** Custom instruction template for the judge */
+  /** Custom judge instruction */
+  instruction: z.string().optional(),
+  /** Deprecated alias */
   instruction_template: z.string().optional(),
   /** Rules for the judge's decision */
   rules: z.array(LoopMonitorRuleSchema).min(1),
diff --git a/src/core/piece/engine/PieceEngine.ts b/src/core/piece/engine/PieceEngine.ts
index 5c11b00..f452ce9 100644
--- a/src/core/piece/engine/PieceEngine.ts
+++ b/src/core/piece/engine/PieceEngine.ts
@@ -462,10 +462,10 @@ export class PieceEngine extends EventEmitter {
   }
 
   /**
-   * Build the default instruction template for a loop monitor judge.
-   * Used when the monitor config does not specify a custom instruction_template.
+   * Build the default instruction for a loop monitor judge.
+   * Used when the monitor config does not specify a custom instruction.
    */
-  private buildDefaultJudgeInstructionTemplate(
+  private buildDefaultJudgeInstruction(
     monitor: LoopMonitorConfig,
     cycleCount: number,
     language: string,
@@ -513,11 +513,11 @@ export class PieceEngine extends EventEmitter {
     cycleCount: number,
   ): Promise<string> {
     const language = this.options.language ?? 'en';
-    const instructionTemplate = monitor.judge.instructionTemplate
-      ?? this.buildDefaultJudgeInstructionTemplate(monitor, cycleCount, language);
+    const instruction = monitor.judge.instruction
+      ?? this.buildDefaultJudgeInstruction(monitor, cycleCount, language);
 
-    // Replace {cycle_count} in custom templates
-    const processedTemplate = instructionTemplate.replace(/\{cycle_count\}/g, String(cycleCount));
+    // Replace {cycle_count} in custom instructions
+    const processedInstruction = instruction.replace(/\{cycle_count\}/g, String(cycleCount));
 
     // Build a synthetic PieceMovement for the judge
     const judgeMovement: PieceMovement = {
@@ -531,7 +531,7 @@ export class PieceEngine extends EventEmitter {
           allowedTools: ['Read', 'Glob', 'Grep'],
         },
       },
-      instructionTemplate: processedTemplate,
+      instruction: processedInstruction,
       rules: monitor.judge.rules.map((r) => ({
         condition: r.condition,
         next: r.next,
@@ -553,7 +553,7 @@ export class PieceEngine extends EventEmitter {
 
     this.emit('movement:start', judgeMovement, this.state.iteration, prebuiltInstruction, this.optionsBuilder.resolveStepProviderModel(judgeMovement));
 
-    const { response, instruction } = await this.movementExecutor.runNormalMovement(
+    const { response, instruction: executedInstruction } = await this.movementExecutor.runNormalMovement(
       judgeMovement,
       this.state,
       this.task,
@@ -562,7 +562,7 @@ export class PieceEngine extends EventEmitter {
       prebuiltInstruction,
     );
     this.emitCollectedReports();
-    this.emit('movement:complete', judgeMovement, response, instruction);
+    this.emit('movement:complete', judgeMovement, response, executedInstruction);
 
     // Resolve next movement from the judge's rules
     const nextMovement = this.resolveNextMovementFromDone(judgeMovement, response);
diff --git a/src/core/piece/engine/team-leader-common.ts b/src/core/piece/engine/team-leader-common.ts
index b34b756..a5352bb 100644
--- a/src/core/piece/engine/team-leader-common.ts
+++ b/src/core/piece/engine/team-leader-common.ts
@@ -45,7 +45,7 @@ export function createPartMovement(step: PieceMovement, part: PartDefinition): P
     model: step.model,
     requiredPermissionMode: step.teamLeader.partPermissionMode ?? step.requiredPermissionMode,
     edit: step.teamLeader.partEdit ?? step.edit,
-    instructionTemplate: part.instruction,
+    instruction: part.instruction,
     passPreviousResponse: false,
   };
 }
diff --git a/src/core/piece/instruction/InstructionBuilder.ts b/src/core/piece/instruction/InstructionBuilder.ts
index 2d9d5e6..e44873a 100644
--- a/src/core/piece/instruction/InstructionBuilder.ts
+++ b/src/core/piece/instruction/InstructionBuilder.ts
@@ -89,7 +89,7 @@ export class InstructionBuilder {
     }
 
     // Skip auto-injection for sections whose placeholders exist in the template
-    const tmpl = this.step.instructionTemplate;
+    const tmpl = this.step.instruction;
     const hasTaskPlaceholder = tmpl.includes('{task}');
     const hasPreviousResponsePlaceholder = tmpl.includes('{previous_response}');
     const hasUserInputsPlaceholder = tmpl.includes('{user_inputs}');
@@ -120,9 +120,9 @@ export class InstructionBuilder {
       ? escapeTemplateChars(this.context.userInputs.join('\n'))
       : '';
 
-    // Instructions (instruction_template processed)
+    // Instructions (movement instruction with placeholder processing)
     const instructions = replaceTemplatePlaceholders(
-      this.step.instructionTemplate,
+      tmpl,
       this.step,
       {
         ...this.context,
diff --git a/src/infra/config/loaders/pieceParser.ts b/src/infra/config/loaders/pieceParser.ts
index 1aecbe7..a7eb4bd 100644
--- a/src/infra/config/loaders/pieceParser.ts
+++ b/src/infra/config/loaders/pieceParser.ts
@@ -272,6 +272,12 @@ function normalizeStepFromRaw(
   const expandedInstruction = step.instruction
     ? resolveRefToContent(step.instruction, sections.resolvedInstructions, pieceDir, 'instructions', context)
     : undefined;
+  if (step.instruction_template !== undefined) {
+    console.warn(`Movement "${step.name}" uses deprecated field "instruction_template". Use "instruction" instead.`);
+  }
+  const expandedLegacyInstruction = step.instruction_template
+    ? resolveRefToContent(step.instruction_template, sections.resolvedInstructions, pieceDir, 'instructions', context)
+    : undefined;
 
   const result: PieceMovement = {
     name: step.name,
@@ -286,9 +292,7 @@ function normalizeStepFromRaw(
     requiredPermissionMode: step.required_permission_mode,
     providerOptions: mergeProviderOptions(inheritedProviderOptions, normalizedProvider.providerOptions),
     edit: step.edit,
-    instructionTemplate: (step.instruction_template
-      ? resolveRefToContent(step.instruction_template, sections.resolvedInstructions, pieceDir, 'instructions', context)
-      : undefined) || expandedInstruction || '{task}',
+    instruction: expandedInstruction || expandedLegacyInstruction || '{task}',
     rules,
     outputContracts: normalizeOutputContracts(step.output_contracts, pieceDir, sections.resolvedReportFormats, context),
     qualityGates: applyQualityGateOverrides(
@@ -335,19 +339,25 @@ function normalizeStepFromRaw(
 
 /** Normalize a raw loop monitor judge from YAML into internal format. */
 function normalizeLoopMonitorJudge(
-  raw: { persona?: string; instruction_template?: string; rules: Array<{ condition: string; next: string }> },
+  raw: { persona?: string; instruction?: string; instruction_template?: string; rules: Array<{ condition: string; next: string }> },
   pieceDir: string,
   sections: PieceSections,
   context?: FacetResolutionContext,
 ): LoopMonitorJudge {
   const { personaSpec, personaPath } = resolvePersona(raw.persona, sections, pieceDir, context);
+  if (raw.instruction_template !== undefined) {
+    console.warn('loop_monitors judge uses deprecated field "instruction_template". Use "instruction" instead.');
+  }
+  const resolvedInstruction = raw.instruction
+    ? resolveRefToContent(raw.instruction, sections.resolvedInstructions, pieceDir, 'instructions', context)
+    : raw.instruction_template
+      ? resolveRefToContent(raw.instruction_template, sections.resolvedInstructions, pieceDir, 'instructions', context)
+      : undefined;
 
   return {
     persona: personaSpec,
     personaPath,
-    instructionTemplate: raw.instruction_template
-      ? resolveRefToContent(raw.instruction_template, sections.resolvedInstructions, pieceDir, 'instructions', context)
-      : undefined,
+    instruction: resolvedInstruction,
     rules: raw.rules.map((r) => ({ condition: r.condition, next: r.next })),
   };
 }
@@ -356,7 +366,7 @@ function normalizeLoopMonitorJudge(
  * Normalize raw loop monitors from YAML into internal format.
  */
 function normalizeLoopMonitors(
-  raw: Array<{ cycle: string[]; threshold: number; judge: { persona?: string; instruction_template?: string; rules: Array<{ condition: string; next: string }> } }> | undefined,
+  raw: Array<{ cycle: string[]; threshold: number; judge: { persona?: string; instruction?: string; instruction_template?: string; rules: Array<{ condition: string; next: string }> } }> | undefined,
   pieceDir: string,
   sections: PieceSections,
   context?: FacetResolutionContext,
diff --git a/src/infra/config/loaders/pieceResolver.ts b/src/infra/config/loaders/pieceResolver.ts
index fab98ee..0628ee5 100644
--- a/src/infra/config/loaders/pieceResolver.ts
+++ b/src/infra/config/loaders/pieceResolver.ts
@@ -229,7 +229,7 @@ function buildMovementPreviews(piece: PieceConfig, maxCount: number): MovementPr
       name: movement.name,
       personaDisplayName: movement.personaDisplayName,
       personaContent: readMovementPersona(movement),
-      instructionContent: movement.instructionTemplate,
+      instructionContent: movement.instruction,
       allowedTools: movement.providerOptions?.claude?.allowedTools ?? [],
       canEdit: movement.edit === true,
     });