From software
Use this skill when the user wants to review code changes, verify implementation quality, or check readiness before a PR or deployment. Trigger on phrases like "review my changes", "run comprehensive review", "check all my code", "am I ready for PR", "before I create a PR", "before I merge", "code audit", "quality check", "verify my implementation", "is my feature complete", "run all reviewers", "check if my implementation is complete", "deep code review", "deep review", or any multi-dimensional code analysis request. Also use proactively when the user says they've finished a phase, completed a feature, or are wrapping up work — even if they don't explicitly ask for a "review". Covers requirements compliance, test coverage, edge cases, architecture alignment, and security.
How this skill is triggered — by the user, by Claude, or both
Slash command
/software:comprehensive-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Orchestrates five specialized review agents in parallel to provide a complete, multi-dimensional analysis of recent code changes:
Orchestrates five specialized review agents in parallel to provide a complete, multi-dimensional analysis of recent code changes:
requirements/, docs/, planning/, or phase plan files (phase-1.md through phase-8.md). If none found, ask the user.git log and git diff to understand scopeLaunch all agents in a single message using multiple Task tool calls. Include specific file paths and context in each prompt so agents don't waste time on discovery.
product-reviewer:
"Review recent changes against [requirements-file]. Map each requirement to its implementation. Flag anything missing or partially implemented. The project uses an 8-phase clean architecture structure."
code-reviewer-business-logic:
"Review recent changes for business logic testability and unit test coverage. Focus on services, use cases, and domain logic. Flag untestable patterns (hard-coded dependencies, global state, non-determinism) and missing unit tests."
code-reviewer-edge-case:
"Review recent changes for unhandled edge cases: null/undefined access, empty collections, boundary values, error scenarios (network timeout, DB failure, API errors), date/time issues, and concurrency problems."
code-reviewer-architecture:
"Review recent changes for clean architecture compliance. Check dependency direction (inner layers must not depend on outer), layer separation (business logic out of UI), and correct phase placement using the 8-phase structure: Phase 1 Foundational → Phase 2 Models → Phase 3 Services → Phase 4 Data → Phase 5 Rules → Phase 6 State → Phase 7 UI → Phase 8 Tests."
code-reviewer-security:
"Review recent changes for security vulnerabilities: injection flaws (SQL, command, template), authentication/authorization issues, sensitive data exposure, hardcoded secrets, weak cryptography (MD5/SHA1/DES), XSS, CSRF, and missing input validation. Reference OWASP Top 10."
Wait for all five agents. Capture full reports and note critical issues.
Use the following structure. Status icons reflect actual findings — do not default to ✅ if there are issues.
# Comprehensive Review Report
## Executive Summary
**Review Scope:**
- Requirements: [which documents]
- Changes: [commit range or files]
- Files Reviewed: [count]
**Overall Status:** [Pass / Pass with Warnings / Needs Attention / Critical Issues]
**Critical Issues:** [count] | **Warnings:** [count] | **Recommendations:** [count]
## Review Dimensions
| Dimension | Status | Summary |
|---|---|---|
| Requirements Compliance | [✅/⚠️/❌] | [X]% implemented, [N] missing |
| Test Coverage | [✅/⚠️/❌] | [X]% coverage, [N] untestable |
| Edge Case Handling | [✅/⚠️/❌] | [N] critical, [N] warning |
| Architecture Alignment | [✅/⚠️/❌] | [N] violations, [N] warnings |
| Security | [✅/⚠️/❌] | [N] critical, [N] high |
**Status icons:** ✅ Good | ⚠️ Needs Attention | ❌ Critical Issues
## Priority Actions
### Must Fix Immediately
1. [Critical issue — source agent, file:line]
### Should Fix Soon
1. [Important issue]
### Consider for Future
1. [Recommendation]
---
## Detailed Reports
### 1. Requirements Compliance
[Full product-reviewer report]
### 2. Test Coverage
[Full code-reviewer-business-logic report]
### 3. Edge Case Analysis
[Full code-reviewer-edge-case report]
### 4. Architecture Review
[Full code-reviewer-architecture report]
### 5. Security Review
[Full code-reviewer-security report]
---
## Next Steps
[Prioritized action list based on combined findings]
No requirements found: Ask the user to provide a requirements file path, master plan, or confirm whether to skip the requirements review.
No recent changes detected:
Ask the user for a commit range (e.g., main..feature-branch) or specific files to review.
An agent fails: Note the failure, continue with the remaining agents, and present a partial report. Suggest running the failed review separately.
references/agent-capabilities.md — Read this when you need to understand what a specific agent analyzes or how to interpret its metrics (e.g., what "critical" vs "warning" means per dimension).references/review-interpretation.md — Read this when consolidating findings: contains decision matrices for go/no-go decisions, cross-cutting patterns (issues appearing in multiple reports), and how to prioritize fixes across dimensions.This skill identifies issues but cannot fix them, modify code, or replace human judgment. Use it as part of your quality process, not as a substitute for it.
npx claudepluginhub hirogakatageri/hirokata --plugin softwareValidates code changes against original plan using git diffs and full file reads, then launches parallel subagents for quality, security, and test coverage checks. Use post-implementation or /recheck.
Orchestrates multi-dimensional code review using specialized agents for code quality, architecture, security, performance, testing, and documentation.
Reviews code changes from MRs/PRs, local commits, or uncommitted files for document compliance, content quality (architecture, tests, observability), and end-to-end consistency (User Story to observability).