Conducting Plan Review

Conducting Plan Review

Overview

Systematic plan evaluation process ensuring plans are comprehensive, executable, and account for all quality criteria (security, testing, architecture, error handling, code quality, process) before implementation begins.

When to Use

Use verifying-plans when:

• Before executing implementation plan: Validate quality and completeness before agents start work
• After writing a plan: Quality-check the plan you just created
• Before high-stakes work: Ensure plan meets standards before committing resources
• When plan scope is uncertain: Verify all requirements are covered
• Default before /execute: Standard quality gate before plan execution

Don't use when:

• Plan is simple checklist (1-3 trivial steps)
• Doing research/exploration (not implementation plan)
• Plan already executed and complete

Quick Reference

Before starting:

1. Read plan to understand scope and approach
2. Evaluate against plan quality standards
3. Check plan structure (task granularity, completeness, TDD approach)
4. Save structured feedback to work directory

Core workflow:

1. Identify plan to review
2. Review against quality checklist (all categories)
3. Evaluate plan structure and completeness
4. Save structured feedback to work directory

Implementation

Prerequisites

Read these to understand quality standards:

• ${CLAUDE_PLUGIN_ROOT}standards/code-review.md - Quality standards apply to plans too
• ${CLAUDE_PLUGIN_ROOT}principles/development.md - Development principles (includes testing)

Step-by-Step Workflow

1. Identify plan to review

Locate the plan:

• Plan files are typically in .work/<feature-name> directory
• Naming pattern: YYYY-MM-DD-<feature-name>.md
• Check current directory or ask user for plan location

Read the plan completely:

• Understand the goal and architecture
• Review all tasks and steps
• Note any immediate concerns

2. Review against quality checklist

Review ALL categories from verify-plan-template.md:

1. Security & Correctness (6 items)

   • Does plan address security vulnerabilities in design?
   • Does plan consider dependency security?
   • Does plan include acceptance criteria?
   • Does plan handle concurrency if applicable?
   • Does plan specify error handling strategy?
   • Does plan address API/schema compatibility?

2. Testing (6 items)

   • Does plan include test strategy?
   • Does plan specify TDD approach?
   • Does plan identify edge cases?
   • Does plan emphasize behavior testing?
   • Does plan require test isolation?
   • Does plan specify test structure?

3. Architecture (7 items)

   • Does plan maintain SRP?
   • Does plan avoid duplication?
   • Does plan separate concerns?
   • Does plan avoid over-engineering (YAGNI)?
   • Does plan minimize coupling?
   • Does plan maintain encapsulation?
   • Does plan keep modules testable?

4. Error Handling (3 items)

   • Does plan specify error handling approach?
   • Does plan include error message requirements?
   • Does plan identify invariants?

5. Code Quality (7 items)

   • Does plan emphasize simplicity?
   • Does plan include naming conventions?
   • Does plan maintain type safety?
   • Does plan follow project patterns?
   • Does plan avoid magic numbers?
   • Does plan specify where rationale is needed?
   • Does plan include documentation requirements?

6. Process (6 items)

   • Does plan include verification steps?
   • Does plan identify performance considerations?
   • Does plan include linting/formatting verification?
   • Does plan scope match requirements?
   • Does plan leverage existing libraries/patterns?
   • Does plan include commit strategy?

Empty BLOCKING section is GOOD if you actually checked. Missing sections mean you didn't check.

BLOCKING vs SUGGESTIONS decision:

Use BLOCKING when:

• Security vulnerability in design
• Missing error handling strategy
• No test strategy or TDD approach
• Tasks too large (>5 minutes)
• Missing exact file paths or commands
• Scope doesn't match requirements

Use SUGGESTIONS when:

• Could add logging for debugging
• Could improve variable naming
• Could add documentation
• Could consider performance optimization
• Could leverage existing pattern

Rule of thumb:

• BLOCKING = Plan will fail during execution or produce insecure/incorrect code
• SUGGESTIONS = Plan would succeed but quality could be higher

3. Evaluate plan structure

Task Granularity:

• Are tasks bite-sized (2-5 minutes each)?
• Are tasks independent where possible?
• Does each task have clear success criteria?

Completeness:

• Are exact file paths specified?
• Are complete code examples provided (not "add validation")?
• Are exact commands with expected output included?
• Are relevant skills/practices referenced?

TDD Approach:

• Does each task follow RED-GREEN-REFACTOR?
• Write test → Run test (fail) → Implement → Run test (pass) → Commit?

4. Save structured evaluation

Template location: ${CLAUDE_PLUGIN_ROOT}templates/verify-plan-template.md

YOU MUST use this exact structure:

# Plan Evaluation - {Date}

## Status: [BLOCKED | APPROVED WITH SUGGESTIONS | APPROVED]

## Plan Summary
- **Feature:** [Feature name]
- **Location:** [Path to plan file]
- **Scope:** [Brief description]

## BLOCKING (Must Address Before Execution)
[Issues or "None"]

**[Issue title]:**
- Description: [what's missing or problematic]
- Impact: [why this blocks execution]
- Action: [what needs to be added/changed]

## SUGGESTIONS (Would Improve Plan Quality)
[Suggestions or "None"]

**[Suggestion title]:**
- Description: [what could be improved]
- Benefit: [how this would help]
- Action: [optional improvement]

## Plan Quality Checklist
[Check all 35 items across 6 categories]

## Plan Structure Quality
[Evaluate task granularity, completeness, TDD approach]

## Assessment
**Ready for execution?** [YES / NO / WITH CHANGES]

**Reasoning:** [Brief explanation]

File naming:

Save to .work/{YYYY-MM-DD}-verify-plan-{HHmmss}.md

Example: .work/2025-11-22-verify-plan-143052.md

Time-based naming ensures:

• No conflicts when multiple agents run in parallel (dual verification)
• Each evaluation gets unique filename automatically
• Collation agents can find all reviews with glob pattern
• No coordination needed between agents

Do NOT create custom section structures. Use template exactly. Additional context (plan excerpts, specific examples) may be added at the end, but core template sections are mandatory.

What NOT to Skip

NEVER skip:

• Reading the entire plan (not just summary)
• Reviewing ALL quality categories (not just critical)
• Checking plan structure (granularity, completeness, TDD)
• Saving evaluation file to work directory
• Including specific examples of issues found

Common rationalizations that violate workflow:

• "Plan looks comprehensive" → Check all categories anyway
• "Author is experienced" → Evaluate objectively regardless of author
• "Just a small feature" → Small features need complete plans
• "Only flagging blockers" → Document suggestions too
• "Template is too detailed" → Template structure is mandatory

Related Skills

Writing plans:

• Writing Plans: ${CLAUDE_PLUGIN_ROOT}skills/writing-plans/SKILL.md

Executing plans:

• Executing Plans: ${CLAUDE_PLUGIN_ROOT}skills/executing-plans/SKILL.md

Testing This Skill

See test-scenarios.md for pressure tests validating this workflow resists rationalization.