diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md
index 8964a81..e208cec 100644
--- a/skills/code-review/SKILL.md
+++ b/skills/code-review/SKILL.md
@@ -1,6 +1,6 @@
 ---
 name: code-review
-description: Review completed implementation work for toki_socket. Reads PLAN.md and CODE_REVIEW.md from the active task, reviews the actual changed source files, appends a verdict to CODE_REVIEW.md, then archives both files as .log. If Required issues are found, writes a new PLAN.md and CODE_REVIEW.md stub so the loop continues immediately.
+description: Review completed implementation work for toki_socket. Reads PLAN.md and CODE_REVIEW.md from the active task, reviews the actual changed source files, appends a verdict to CODE_REVIEW.md, then archives both files as .log. If any issues are found (WARN or FAIL), writes a new PLAN.md and CODE_REVIEW.md stub so the loop continues immediately.
 ---
 
 # Code Review
@@ -29,7 +29,7 @@ Note: `PLAN.md` and `CODE_REVIEW.md` are always present together. The implementi
 ## Step 1 — Find the Active Task
 
 1. Glob `tasks/*/CODE_REVIEW.md`.
-2. If exactly one exists: that is the active task. Record `task_name`. Confirm `PLAN.md` is also present in the same directory.
+2. If exactly one exists: that is the active task. Record `task_name`. `PLAN.md` is always present alongside `CODE_REVIEW.md` — this is an invariant guaranteed by the plan skill and code-review skill. No existence check needed.
 3. If no `CODE_REVIEW.md` found: nothing to review. Stop and report.
 4. If multiple: list them and ask the user which to review. Stop.
 
@@ -53,39 +53,48 @@ No git narrowing. Read everything relevant to the plan:
 2. `tasks/{task_name}/PLAN.md`
 3. Every source file listed in the plan's "수정 파일 및 체크리스트" — read the whole file.
 4. Every test file that exercises the changed code.
-5. Every file that imports or is imported by the changed files.
+5. Files that import or are imported by the changed files — **up to 2 levels deep**.
 
 ### 2-2B. Follow-up review — diff-anchored + expansion
 
-Start with git to identify what changed, then expand outward to connected code.
+Start with git unstaged changes to identify what changed, then expand outward to connected code.
+
+The implementing agent does not commit — changes are always in the working tree (unstaged or staged).
 
 **Collect the diff:**
 ```bash
-git rev-parse HEAD 2>/dev/null && git diff HEAD || (git diff && git diff --cached)
-git diff main..HEAD    # if on a feature branch
-git log --oneline -10
+git diff            # unstaged changes
+git diff --cached   # staged changes
+git log --oneline -5
 ```
 
+Use the union of both as the change set.
+
 **Expand beyond the diff** — for each changed file or symbol, also read:
-- Files that **import** the changed file (callers may be broken by signature changes)
+- Files that **import** the changed file — **up to 2 levels deep** (callers may be broken by signature changes)
 - Files that **implement or extend** a changed interface or abstract class
 - Test files that **exercise** the changed logic, even if not modified themselves
 - Any file the plan's "수정 파일 및 체크리스트" lists that does not appear in the diff (planned but possibly missing)
 
 The diff scopes where to start; it does not limit where to stop. Follow the connections.
 
+> Note: import expansion boundary follows the project's structural rules (e.g., DDD layer rules). If a separate rules file defines allowed dependency directions, use it as the ceiling for how far to trace.
+
 ---
 
 ## Step 3 — Pre-Review Checklist
 
-Complete every item before writing the verdict. Use the git diff as the primary reference — only reach outside it when a specific check demands it.
+Complete every item before writing the verdict.
 
-- [ ] Compare git diff against the plan's "수정 파일 및 체크리스트" — are all planned changes present? Are there unplanned changes?
-- [ ] For every symbol renamed or removed in the diff: `git grep <old_name>` on the pre-change tree (or Grep on current tree) to confirm no stale references remain outside the diff.
-- [ ] For every new test added in the diff: confirm the test name matches the plan, and the assertion is non-trivial (not just "no exception thrown").
-- [ ] Scan the diff for code quality issues: debug prints, dead code, leftover TODOs, commented-out blocks.
-- [ ] Check for unintended diff noise: formatting-only changes, unrelated file edits, accidentally committed debug code.
-- [ ] Cross-check the claimed `dart analyze` / `go test` output in CODE_REVIEW.md against what the diff actually introduces.
+**Applies to all reviews:**
+- [ ] Compare actual source files against the plan's "수정 파일 및 체크리스트" — are all planned changes present in the code?
+- [ ] For every symbol renamed or removed: Grep on the current working tree to confirm no stale references remain.
+- [ ] For every new test listed in the plan: confirm it exists, name matches, assertion is non-trivial.
+- [ ] Cross-check the claimed `dart analyze` / `go test` output in CODE_REVIEW.md against the actual code.
+
+**Follow-up review only** (count ≥ 1, diff available):
+- [ ] Compare `git diff` + `git diff --cached` against the plan's checklist — are there unplanned changes or missing changes?
+- [ ] Scan the diff for noise: debug prints, dead code, leftover TODOs, formatting-only changes, unrelated edits.
 
 ---
 
@@ -125,7 +134,8 @@ Append the following section to the existing `CODE_REVIEW.md`. Every field is re
 
 <!-- 해당하는 하나만 남기고 나머지 줄은 삭제 -->
 **[PASS]** 추가 작업 불필요. 아카이브 완료.
-**[FIX]** Required 문제 N건. 새 PLAN.md 작성 완료. 구현 에이전트는 PLAN.md를 읽고 작업 시작.
+**[WARN]** Suggested/Nit N건. 새 PLAN.md 작성 완료. 구현 에이전트는 PLAN.md를 읽고 작업 시작.
+**[FAIL]** Required N건. 새 PLAN.md 작성 완료. 구현 에이전트는 PLAN.md를 읽고 작업 시작.
 ```
 
 ---
@@ -148,14 +158,16 @@ After this step neither `PLAN.md` nor `CODE_REVIEW.md` exists.
 
 ## Step 6 — Post-Review Actions
 
-### If verdict is PASS (zero Required issues)
+### If verdict is PASS (zero issues)
 
 No further files needed. Report to the user:
 - Verdict: PASS
 - Archives: `code_review_N.log`, `plan_M.log`
 - Task status: complete (only `.log` files remain in `tasks/{task_name}/`)
 
-### If verdict is WARN or FAIL (one or more Required issues)
+### If verdict is WARN or FAIL
+
+WARN = Suggested/Nit issues only, zero Required. FAIL = one or more Required issues. Both unconditionally write new PLAN.md and CODE_REVIEW.md stub.
 
 Write new files for the fix round:
 
@@ -169,7 +181,7 @@ Write new files for the fix round:
   - **수정 파일 및 체크리스트**: exhaustive checkbox list.
   - **테스트 작성**: decision + justification.
   - **중간 검증**: runnable command.
-- Suggested/Nit issues may be grouped into one optional item.
+- If verdict is WARN (Suggested/Nit issues, no Required): group all Suggested and Nit issues into one plan item. Include it in the plan — it is not optional.
 - Include 수정 파일 요약 table.
 - Include 최종 검증 section.
 
@@ -179,7 +191,7 @@ Write new files for the fix round:
 - Empty sections: 계획 대비 변경 사항, 주요 설계 결정, 리뷰어 체크포인트, 검증 결과.
 
 Report to the user:
-- Verdict: WARN/FAIL, N Required issues
+- Verdict: FAIL → Required N건 / WARN → Suggested/Nit N건
 - Archives: `code_review_N.log`, `plan_M.log`
 - New plan: `tasks/{task_name}/PLAN.md` ready for implementing agent
 
@@ -220,5 +232,5 @@ Report to the user:
 2. `plan_M.log` exists (was PLAN.md).
 3. Neither `CODE_REVIEW.md` nor `PLAN.md` exists as `.md` in the task directory.
 4. If PASS: only `.log` files remain in `tasks/{task_name}/`.
-5. If FAIL: new `PLAN.md` exists, header `plan=N` matches count of `plan_*.log` after archiving, covers every Required issue with full sub-sections.
-6. If FAIL: new `CODE_REVIEW.md` stub exists, header `plan=N` matches new PLAN.md, completion table lists every item.
+5. If WARN or FAIL: new `PLAN.md` exists, header `plan=N` matches count of `plan_*.log` after archiving. FAIL: every Required issue has its own plan item with full sub-sections. WARN: grouped Suggested/Nit plan item exists with full sub-sections.
+6. If WARN or FAIL: new `CODE_REVIEW.md` stub exists, header `plan=N` matches new PLAN.md, completion table lists every item.
diff --git a/skills/plan/SKILL.md b/skills/plan/SKILL.md
index 8877890..19ddd58 100644
--- a/skills/plan/SKILL.md
+++ b/skills/plan/SKILL.md
@@ -10,7 +10,7 @@ description: Analyze the codebase and write a detailed PLAN.md for toki_socket i
 This skill produces two files that drive the implementation loop:
 
 ```
-implement skill (this)         → PLAN.md + CODE_REVIEW.md stub
+plan skill (this)              → PLAN.md + CODE_REVIEW.md stub
     ↓
 Separate implementing agent    → reads PLAN.md, codes, fills CODE_REVIEW.md
     ↓
@@ -21,7 +21,7 @@ All files live under `tasks/{task_name}/`. Directory state at any point:
 
 | State | Meaning |
 |-------|---------|
-| `PLAN.md` only (no `CODE_REVIEW.md`) | Should not occur — implement skill always writes both |
+| `PLAN.md` only (no `CODE_REVIEW.md`) | Should not occur — plan skill always writes both |
 | `PLAN.md` + `CODE_REVIEW.md` (stub, unfilled) | Implementing agent is working |
 | `PLAN.md` + `CODE_REVIEW.md` (filled) | Ready for code-review skill |
 | Only `*.log` files | Task complete |
@@ -30,16 +30,17 @@ All files live under `tasks/{task_name}/`. Directory state at any point:
 
 ## Step 1 — Determine Task
 
-### New task
+If the user names the task explicitly, use that name and skip auto-detection.
 
-1. Derive `task_name`: short snake_case from the user request (e.g., `dart_api_refactor`, `go_base_client`).
-2. Create `tasks/{task_name}/` if it does not exist.
+Otherwise, glob `tasks/*/PLAN.md` and branch:
 
-### Continuing an existing task (re-plan triggered by code-review)
+| Result | Action |
+|--------|--------|
+| Exactly one found | Continuing task — use that directory. `task_name` = directory name. |
+| None found | New task — derive `task_name` from the user request (short snake_case, e.g. `dart_api_refactor`). Create `tasks/{task_name}/`. |
+| Multiple found | Cannot auto-detect — list the found paths and ask the user which task to work on. Stop. |
 
-1. Glob `tasks/*/PLAN.md` — if exactly one exists, that is the active task. Use its directory.
-2. If none but `tasks/*/CODE_REVIEW.md` exists: the task is waiting for `code-review`, not this skill. Stop and report.
-3. If the user names the task explicitly, use that name regardless.
+> PLAN.md is the entry point of the loop. Without it the loop has not started.
 
 ---
 
@@ -73,12 +74,15 @@ Before writing new files, archive any active files from a previous round:
 
 ## Step 4 — Write PLAN.md
 
+**N for the header** = count of `tasks/{task_name}/plan_*.log` files after Step 3 (i.e., including the one just archived, if any).
+
 Write `tasks/{task_name}/PLAN.md`:
 
-```markdown
-# <Title>
+> Note: Include the "의존 관계 및 구현 순서" section only when plan items must be done in a specific order. Omit it entirely when items are independent.
 
+```markdown
 <!-- task={task_name} plan=N tag=<TAG> -->
+# <Title>
 
 ## 이 파일을 읽는 구현 에이전트에게
 
@@ -148,7 +152,7 @@ Expected: <N개 통과>
 
 ---
 
-## 의존 관계 및 구현 순서  ← 순서가 있을 때만 포함
+## 의존 관계 및 구현 순서
 
 \```
 [TAG-1] → [TAG-2] → [TAG-3]
@@ -244,7 +248,7 @@ task={task_name}, plan=N, tag=TAG
 
 After writing both files:
 
-1. `tasks/{task_name}/PLAN.md` exists and line 1 is `<!-- task={task_name} plan=N tag=TAG -->`.
+1. `tasks/{task_name}/PLAN.md` exists and **line 1** is `<!-- task={task_name} plan=N tag=TAG -->`.
 2. `tasks/{task_name}/CODE_REVIEW.md` exists and line 1 matches `<!-- task={task_name} plan=N tag=TAG -->`.
 3. Previous `PLAN.md` (if existed) was archived as `plan_N.log` with N = count of pre-existing logs.
 4. Previous `CODE_REVIEW.md` (if existed) was archived as `code_review_M.log`.