home / skills / dexploarer / hyper-forge / verify-claude-setup

verify-claude-setup skill

/.claude/skills/verify-claude-setup

This skill verifies .claude setup completeness, ensuring agents, hooks, rules, memory, and settings are properly configured before work.

npx playbooks add skill dexploarer/hyper-forge --skill verify-claude-setup

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
2.9 KB
---
name: verify-claude-setup
description: Verify .claude directory configuration is complete and correct. Use when checking if agents, hooks, rules, and memory are properly set up, or after making changes to .claude configuration.
allowed-tools: [Read, Bash, Grep, Glob]
---

# Verify .claude Setup

Quick verification that .claude directory is properly configured.

## When to Use

- After updating agents, rules, hooks, or memory
- Before starting complex work (ensure system is ready)
- User asks "is everything set up correctly?"
- Troubleshooting configuration issues

## Verification Checklist

### 1. Agents (should have 6)
```bash
ls .claude/agents/ | wc -l
```

Verify all agents have:
- Proper YAML frontmatter
- Research-First Protocol section
- Clear responsibilities

### 2. Hooks (should have 11 scripts + README)
```bash
ls .claude/hooks/scripts/*.ts | wc -l
```

Check:
- All scripts executable (`chmod +x`)
- All have Bun shebang (`#!/usr/bin/env bun`)
- settings.json has all hooks configured
- Pre-tool and Post-tool hooks exist

### 3. Rules (should have research-first-protocol.md)
```bash
ls .claude/rules/*.md
```

Verify:
- research-first-protocol.md exists
- Has `alwaysApply: true`

### 4. Memory (should have 8 files)
```bash
ls .claude/memory/ | wc -l
```

Check CLAUDE.md imports all:
- research-first-enforcement.md
- coding-standards.md
- testing-standards.md
- architecture-patterns.md
- common-workflows.md
- build-commands.md
- asset-forge-guide.md
- security-protocols.md

### 5. Settings
```bash
grep -c "hooks" .claude/settings.json
```

Verify:
- SessionStart, UserPromptSubmit hooks
- PreToolUse (Bash, Write, Edit) hooks
- PostToolUse (Write, Edit, Read, Grep, Glob) hooks
- PreCompact, Stop, SubagentStop, SessionEnd hooks

### 6. Commands (should have ~20)
```bash
find .claude/commands -type f -name "*.md" | wc -l
```

## Quick Verification

Run all checks:
```bash
echo "Agents: $(ls .claude/agents/ | wc -l)"
echo "Hooks: $(ls .claude/hooks/scripts/*.ts | wc -l)"
echo "Rules: $(ls .claude/rules/*.md | wc -l)"
echo "Memory: $(ls .claude/memory/ | wc -l)"
echo "Commands: $(find .claude/commands -type f -name "*.md" | wc -l)"
echo "Skills: $(ls .claude/skills/ | wc -l)"
```

## Expected Counts

- Agents: 6
- Hook Scripts: 11
- Rules: 2 (README + research-first-protocol)
- Memory: 8
- Commands: ~20
- Skills: ~40

## Report Format

Provide summary:
```
✅ Agents: 6/6 with Research-First Protocol
✅ Hooks: 11/11 executable, all configured
✅ Rules: research-first-protocol.md active
✅ Memory: 8/8 files, all imported in CLAUDE.md
✅ Commands: 20 organized slash commands
✅ Skills: 40 skills ready

Status: .claude configuration VERIFIED
```

Or if issues found:
```
⚠️ Agents: Missing Research-First Protocol in database-specialist.md
⚠️ Hooks: pre-tool-write.ts not executable
❌ Memory: research-first-enforcement.md not imported in CLAUDE.md

Status: Configuration INCOMPLETE - needs fixes
```

Overview

This skill verifies that the .claude directory configuration is complete and correct for the project. It runs a checklist validating agents, hooks, rules, memory imports, settings, commands, and skills to confirm everything needed for reliable agent operation is present. Use it to catch configuration gaps after edits or before critical runs.

How this skill works

The skill inspects file counts and specific file contents in .claude subfolders, ensuring expected items exist and are properly configured. It checks agent frontmatter and protocol sections, hook script counts and shebangs/permissions, rule activation flags, memory file imports in CLAUDE.md, relevant settings entries, and the approximate number of commands and skills. It returns a concise pass/fail summary and highlights specific missing or misconfigured items.

When to use it

  • After updating agents, hooks, rules, or memory files
  • Before starting complex or long-running agent workflows
  • When the system reports unexpected agent behavior
  • During onboarding or repository audits
  • Before a release or deployment that relies on automated agents

Best practices

  • Run the verification after every change to .claude to catch regressions early
  • Keep agent YAML frontmatter and Research-First Protocol sections consistent
  • Ensure hook scripts have the Bun shebang and executable permissions
  • Maintain CLAUDE.md as the single source of truth for memory imports
  • Version-control settings.json and validate hook lists after edits

Example use cases

  • Confirm all six agents are present and include a Research-First Protocol before starting a multi-agent run
  • Detect a non-executable hook script after adding new hooks to .claude/hooks/scripts
  • Verify that CLAUDE.md imports all eight memory files after reorganizing docs
  • Quick pre-release check to ensure hook and rule configurations are intact

FAQ

What does the skill actually count?

It counts agents, hook scripts, rule files, memory files, command documents, and skills to compare against expected totals and flags missing or extra items.

Will it modify any files?

No. The skill performs read-only checks and reports findings; remediation steps are suggested but not applied automatically.