# Sdd Review Plan

> sdd-review-plan

- Skill: `sc30gsw/sdd-review-plan` (Agent Skill)
- Install (CLI): `npx skillmds@latest add sc30gsw/sdd-review-plan`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sc30gsw/sdd-review-plan/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: sc30gsw (https://skillmd.com/u/sc30gsw)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/sc30gsw/sdd-review-plan

---

# sdd-review-plan

**Slash command**: `/sdd-review-plan <slug>`
**Purpose**: Comprehensive pre-implementation review of requirements, design, and tasks. Writes "Traceability Coherence" and "Plan Review" findings to `review-results/plan-review.md`.

---

## Prerequisites

- `.claude/specs/<slug>/requirements.md` must exist
- `.claude/specs/<slug>/design.md` must exist
- `.claude/specs/<slug>/tasks.md` must exist
- `.claude/specs/<slug>/progress.md` must exist (used to read `mode`)
- `.claude/specs/<slug>/review-results/requirement-review.md` (created by sdd-review-requirements)

---

## Steps

### 1. Read all spec inputs

```
.claude/specs/<slug>/requirements.md
.claude/specs/<slug>/design.md
.claude/specs/<slug>/tasks.md
.claude/specs/<slug>/progress.md    (read mode: standard | auto)
.claude/specs/<slug>/review-results/requirement-review.md  (for prior review context)
```

Build three indexes in memory:

- **REQ index**: all `REQ-XXX` IDs found in requirements.md
- **Design section index**: all `§X.X <title>` headings found in design.md
- **TASK index**: all `TASK-XXX` IDs found in tasks.md

### 2. Run traceability coherence check (BOTH modes)

Run all six checks below. For each check, record either ✅ (pass) or ❌ (fail) with specific IDs.

#### Check A — REQ → Design coverage

Every REQ-XXX in the REQ index appears in at least one `Satisfies:` line in design.md.

Failure example:

```
❌ REQ-004 has no Satisfies: line in design.md
   Fix: add "Satisfies: REQ-004" to the relevant design section
```

#### Check B — REQ → Task coverage

Every REQ-XXX in the REQ index appears in at least one `Implements:` line in tasks.md.

Failure example:

```
❌ REQ-006 has no Implements: entry in tasks.md
   Fix: add REQ-006 to an existing task's Implements field, or create TASK-NNN
```

#### Check C — Design → Task coverage

Every design section §X.X in the design section index appears in at least one `Design ref:` line in tasks.md.

Failure example:

```
❌ §7 Error Handling Strategy has no Design ref: entry in tasks.md
   Fix: add "Design ref: §7 Error Handling Strategy" to the relevant task
```

#### Check D — Task → REQ completeness

Every TASK-XXX in tasks.md has at least one `Implements:` entry (not empty).

Failure example:

```
❌ TASK-005 has no Implements: field
   Fix: add "Implements: REQ-XXX" to TASK-005
```

#### Check E — Dangling references

All REQ-XXX IDs referenced in `Satisfies:` or `Implements:` fields actually exist in the REQ index.
All §X.X references in `Design ref:` fields actually exist in the design section index.

Failure example:

```
❌ tasks.md TASK-003 references "Design ref: §9 Deployment" but §9 does not exist in design.md
   Fix: correct the section reference or add §9 to design.md
```

#### Check F — Duplicate IDs

No repeated REQ-XXX IDs in requirements.md.
No repeated TASK-XXX IDs in tasks.md.

Failure example:

```
❌ TASK-002 appears twice in tasks.md
   Fix: renumber the second occurrence
```

### 3. Write Traceability Coherence table to `review-results/plan-review.md`

Write to `.claude/specs/<slug>/review-results/plan-review.md`:

```markdown
# Plan Review: <slug>

**Date**: <YYYY-MM-DD>
**Reviewer**: sdd-review-plan skill
**Mode**: <standard|auto>
**Traceability**: ✅ All checks passed (or ⚠️ N failures)

## Traceability Coherence

| Check            | Result  | Details                               |
| ---------------- | ------- | ------------------------------------- |
| A: REQ → Design  | ✅ / ❌ | <!-- IDs or "all covered" -->         |
| B: REQ → Task    | ✅ / ❌ | <!-- IDs or "all covered" -->         |
| C: Design → Task | ✅ / ❌ | <!-- sections or "all covered" -->    |
| D: Task → REQ    | ✅ / ❌ | <!-- IDs or "all have Implements" --> |
| E: Dangling refs | ✅ / ❌ | <!-- broken refs or "none" -->        |
| F: Duplicate IDs | ✅ / ❌ | <!-- duplicates or "none" -->         |
```

### 4. CRITICAL gate — stop on any traceability failure

If ANY check is ❌, output the following and STOP. Do NOT output the PHASE COMPLETE gate.

```
⚠️ TRACEABILITY ERRORS FOUND — cannot proceed to implementation

The following issues must be fixed in the spec files before proceeding:

[list each ❌ with the specific IDs and how to fix]

Fix the issues listed above, then re-run /sdd-review-plan <slug>.
```

Only continue to Step 5 if ALL six checks are ✅.

### 5. Mode-specific plan review

**`--mode standard`**:

1. Invoke `ecc:planner` agent with the full requirements + design + tasks context
2. Invoke `ecc:architect` agent with the same context
3. Invoke the `ecc:docs-lookup` agent for the technology stack used in design.md
4. Collect all findings

**`--mode auto`**:

1. Invoke `ecc:architect` agent only (skip ecc:planner agent and ecc:docs-lookup agent)
2. Collect findings

### 6. Write Plan Review to `review-results/plan-review.md`

Write findings to plan-review.md after the Traceability Coherence section:

```markdown
## Architecture & Design Review

| Section / 横断 | Issue | Severity | Suggestion |
| -------------- | ----- | -------- | ---------- |
| ...            | ...   | ...      | ...        |

## Task Breakdown Review

| Task / 横断 | Issue | Severity | Suggestion |
| ----------- | ----- | -------- | ---------- |
| ...         | ...   | ...      | ...        |

## Technology Stack Review (standard mode only)

| 確認項目 | 結果 | 影響 |
| -------- | ---- | ---- |
| ...      | ...  | ...  |

## Recommended Actions Before Proceeding

- [ ] **HIGH**: <action 1>
- [ ] **HIGH**: <action 2>
- [ ] **MEDIUM**: <action 3>

## Sign-off Condition

All HIGH severity findings must be resolved before running `/sdd-impl <slug>`.
```

---

## Output

```
.claude/specs/<slug>/review-results/plan-review.md    (created or overwritten)
```

---

## change-log.md Update

After writing plan-review.md, append to `.claude/specs/<slug>/change-log.md`:

```
| <YYYY-MM-DD> | sdd-review-plan | plan-review.md 作成 (Traceability: ✅/⚠️, <N> findings) |
```

---

## Phase Gate

Only output this block when ALL six traceability checks are ✅ AND plan review is complete:

```
== PHASE COMPLETE: sdd-review-plan ==
Artifact: .claude/specs/<slug>/review-results/plan-review.md
Summary:
- All 6 traceability coherence checks passed (A-F)
- Plan reviewed by ecc:architect agent (+ ecc:planner agent + ecc:docs-lookup agent in standard mode)
- Key risks and open questions documented in review-results/plan-review.md
- Spec is ready for implementation
- Run /sdd-impl <slug> TASK-001 to begin

⏸ WAITING FOR CONFIRMATION
Type `CONFIRM sdd-impl` to proceed with implementation, or describe changes needed.
```

