validate-agent

​

Syntax

vg validate-agent <path>

​

Description

• Script security validation
• Agent instruction files (.cursorrules, .clinerules, etc.)
• Common agent patterns and idioms
• Batch validation of multiple scripts

• Validating agent-generated scripts before execution
• Auditing agent instruction files
• Ensuring agent workflows are secure
• CI/CD validation of agent outputs

​

Arguments

​

Examples

​

Validate Agent Script

vg validate-agent agent-generated-script.sh

✅ PASS: agent-generated-script.sh
   No security issues detected

Summary:
- Total scripts: 1
- Passed: 1
- Failed: 0

​

Validate Agent Directory

# Validate all scripts in a directory
vg validate-agent scripts/agent-output/

Validating scripts in: scripts/agent-output/

✅ PASS: setup.sh
⚠️  WARN: deploy.sh
   - Medium risk: sudo usage detected
   - Line 15: sudo systemctl restart nginx

❌ FAIL: cleanup.sh
   - High risk: rm -rf with variable expansion
   - Line 8: rm -rf $TEMP_DIR/*

Summary:
- Total scripts: 3
- Passed: 1
- Warnings: 1
- Failed: 1

​

Validate Agent Instructions

# Check agent rule files
vg validate-agent .cursorrules
vg validate-agent .clinerules

​

CI/CD Integration

# Validate all agent-generated scripts in CI
vg validate-agent ci/agent-scripts/ || {
  echo "Agent script validation failed"
  exit 1
}

​

Validation Checks

1. Shell Script Analysis
   • Syntax validation
   • Security pattern detection
   • Risk level assessment
2. Agent-Specific Checks
   • Agent instruction file syntax
   • Common agent antipatterns
   • Agent-generated code patterns
3. Directory Traversal
   • Recursive script discovery
   • Batch validation
   • Aggregate reporting

​

Exit Codes

• 0: All scripts passed validation
• 1: One or more scripts failed validation
• 2: Validation error (file not found, parse error)

​

Supported Agent Files

• .cursorrules - Cursor IDE rules
• .clinerules - CLI agent rules
• .aiderules - Aider agent rules
• .windsurfrules - Windsurf agent rules
• Any .sh, .bash, .zsh scripts

​

Use Cases

​

Pre-Execution Validation

# Agent generates script
agent_output="generated-deploy.sh"

# Validate before running
vg validate-agent "$agent_output"
if [ $? -eq 0 ]; then
  # Safe to execute
  vg exec -- bash "$agent_output"
else
  echo "Agent script validation failed - manual review required"
fi

​

Batch Agent Output Validation

# Validate all scripts generated in a session
SESSION_DIR=".agent-session-$(date +%s)"

# After agent completes
vg validate-agent "$SESSION_DIR"

​

IDE Integration

# Add to IDE task
{
  "label": "Validate Agent Scripts",
  "type": "shell",
  "command": "vg validate-agent scripts/",
  "problemMatcher": []
}

​

Agent-Specific Patterns

• Script generation templates
• Agent command wrappers
• Temporary file management
• Agent-specific error handling

​

Related Commands

• validate - Validate individual scripts
• exec - Execute validated scripts
• prompt-firewall - Check agent prompts for injection
• scan-security - Deep security scan