---
title: "/ork:verify"
description: "Comprehensive verification with 5 parallel agents, nuanced 0-10 grading, and prioritized improvement suggestions."
canonical: "https://orchestkit.yonyon.ai/docs/skills/spotlights/verify"
---

# /ork:verify

Comprehensive verification with 5 parallel agents, nuanced 0-10 grading, and prioritized improvement suggestions.

import { Callout } from 'fumadocs-ui/components/callout';

The `/ork:verify` command gives you an honest, structured assessment of your implementation. Five specialized agents evaluate code quality, security, test coverage, API compliance, and UI patterns simultaneously. The results are combined into a weighted composite score (0-10) with a letter grade and a prioritized list of improvements ranked by effort vs. impact. This page explains the grading system, how to interpret results, and when to use verify in your workflow.

## Why This Skill

Code review catches issues after you think you are done. Verification catches them before you commit. The difference matters: fixing a security vulnerability before commit is a 5-minute edit; fixing it after review means a new commit, re-review, and potentially a rebase. `/ork:verify` sits between implementation and commit, giving you a quality report while the code is still fresh in your working tree.

It is also useful as a standalone audit. Run it on code you did not write -- a new dependency, a colleague's branch, or legacy code you are about to modify -- and get a structured assessment before you touch anything.

## Quick Start

```bash
/ork:verify authentication flow
/ork:verify user profile feature
/ork:verify --scope=backend database migrations
```

After invocation, OrchestKit asks for the verification scope:

- **Full verification**: All 5 agents, complete grading and suggestions
- **Tests only**: Run unit + integration + e2e tests, skip analysis
- **Security audit**: Focus on vulnerabilities
- **Code quality**: Focus on lint, types, complexity
- **Quick check**: Run tests only, skip detailed scoring

## How It Works

### Phase 1: Context Gathering

OrchestKit examines what changed since the base branch:

```bash
git diff main --stat
git log main..HEAD --oneline
git diff main --name-only
```

This scopes the verification to relevant files rather than auditing the entire codebase.

### Phase 2: Parallel Agent Dispatch (5 Agents)

Five agents run simultaneously, each scoring their dimension 0-10:

| Agent | Dimension | What It Checks |
|-------|-----------|----------------|
| `code-quality-reviewer` | Code Quality | Naming, complexity, DRY, SOLID principles |
| `security-auditor` | Security | OWASP, secrets, injection, dependencies |
| `test-generator` | Test Coverage | Coverage %, test quality, edge cases, flaky tests |
| `backend-system-architect` | API Compliance | REST conventions, async patterns, N+1 queries |
| `frontend-ui-developer` | UI Compliance | React 19 patterns, accessibility, loading states |

### Phase 3: Composite Scoring

The individual scores are weighted into a composite:

| Dimension | Weight |
|-----------|--------|
| Code Quality | 20% |
| Security | 25% |
| Test Coverage | 20% |
| API Compliance | 20% |
| UI Compliance | 15% |

Security has the highest weight because security issues are the most expensive to fix after deployment.

### Grade Interpretation

| Score | Grade | What It Means |
|-------|-------|---------------|
| 9.0-10.0 | A+ | Ship it. No improvements needed. |
| 8.0-8.9 | A | Ready for merge. Minor polish optional. |
| 7.0-7.9 | B | Solid work. A few improvements would elevate it. |
| 6.0-6.9 | C | Functional but rough. Consider improvements before merge. |
| 5.0-5.9 | D | Significant issues. Improvements strongly recommended. |
| 0.0-4.9 | F | Do not merge. Fundamental problems to address. |

### Phase 4: Improvement Suggestions

Each suggestion includes an effort estimate (1-5) and impact rating (1-5). Priority is calculated as `impact / effort`, surfacing quick wins first:

| Points | Effort | Impact |
|--------|--------|--------|
| 1 | < 15 min | Minimal |
| 2 | 15-60 min | Low |
| 3 | 1-4 hours | Medium |
| 4 | 4-8 hours | High |
| 5 | 1+ days | Critical |

**Quick wins** are suggestions with effort 2 or below and impact 4 or above. These are the first things to address because they deliver the most value for the least work.

## Common Patterns

### Pattern 1: Post-Implementation Verification

The most common use -- verify your own work before committing:

```bash
/ork:implement search feature
# ... implementation complete ...
/ork:verify search feature
/ork:commit
```

### Pattern 2: Pre-Review Audit

Before reviewing someone else's PR, run verify on their branch:

```bash
git checkout feature/user-auth
/ork:verify --scope=backend
```

This gives you a structured report before you start reading code, so you know where to focus your attention.

### Pattern 3: Security-Focused Check

When the change touches authentication, authorization, or data handling:

```bash
/ork:verify --scope=security payment processing
```

Only the `security-auditor` agent runs, but it runs deeper: checking for OWASP Top 10, secrets in code, dependency vulnerabilities, and injection vectors.

### Pattern 4: Continuous Quality Tracking

The verify skill stores scores in memory. Over time, you can query trends:

```bash
/ork:memory search "VerificationMetrics"
```

This shows how your codebase quality evolves across implementations.

## Tips and Tricks

<Callout type="info">
**Run verify before every PR.** Making `/ork:verify` a habit catches issues early. A 30-second verification pass is cheaper than a review round-trip that takes hours because a security issue was missed.
</Callout>

<Callout type="warn">
**A score below 5.0 on security is a hard blocker.** The default policy blocks merges when the security dimension scores below 5.0. This is intentional -- security issues compound, and shipping a known vulnerability creates technical debt that is expensive to remediate.
</Callout>

<Callout type="info">
**Quick wins first.** The improvement suggestions are sorted by priority (impact/effort). Start with quick wins: they are fast to implement and have outsized impact on the composite score. A 15-minute fix that raises your security score from 6 to 8 is always worth doing.
</Callout>

<Callout type="info">
**Use Policy-as-Code to customize thresholds.** Create `.claude/policies/verification-policy.json` to set minimum scores per dimension, blocking rules, and coverage targets. This lets your team enforce standards automatically rather than relying on manual review.
</Callout>

## Related Skills

- [/ork:implement](/docs/skills/spotlights/implement) -- Build features with built-in quality gates
- [/ork:review-pr](/docs/skills/spotlights/review-pr) -- Review pull requests with parallel agents
- [/ork:commit](/docs/skills/spotlights/commit) -- Commit verified changes
- [Command Skills Reference](/docs/skills/command-skills) -- All 35 command skills