Ai Runbooks

Agentic SOC Runbooks

This repository provides security operations runbooks and role-based guides for AI-assisted cybersecurity workflows.

Overview

Built by and for enterprise security teams, this project standardizes incident response procedures and threat analysis workflows across multiple AI platforms including Claude, Gemini, and Cline. It equips LLMs with specialized context and cognitive tools to perform complex security investigations with the same rigor and consistency as experienced analysts.

Key Components

Personas

Security role definitions including:

• SOC Analysts (Tier 1-3)
• Threat Hunters
• Incident Responders
• CTI Researchers
• Security Engineers
• And more...

Runbooks

Step-by-step procedural guides for:

• Alert triage
• IOC enrichment and investigation
• Threat hunting
• Incident response
• Case management
• And many more security operations tasks

Incident Response Plans (IRPs)

Comprehensive end-to-end strategies for:

• Malware incidents
• Phishing attacks
• Ransomware
• Compromised accounts

Repository Structure

ai_runbooks/
├── rules_bank/                    # Master source directory for all content
│   ├── personas/                  # Security role definitions
│   ├── run_books/                 # Procedural guides for security tasks
│   │   ├── common_steps/         # Reusable procedure components
│   │   ├── irps/                 # Incident Response Plans
│   │   └── guidelines/           # Best practices and documentation
│   └── [other documentation]     # Configuration and reference files
├── reports/                      # Generated security reports
├── mcp-security/                 # MCP server implementations
└── [configuration files]         # Various setup and config scripts

Usage

1. For Security Teams: Use this repository to standardize and document your security operations procedures for AI-assisted workflows.

1. For Developers:

• Edit content only in the rules_bank/ directory
• Never edit files through the symlinks in AI tool directories
• Use the provided Python scripts for configuration if needed

1. For AI Assistants: The symlinked rules_bank directory in each tool's configuration folder provides access to all personas, runbooks, and documentation.

Integration

This project is designed to work with:

• Chronicle SIEM
• SOAR platforms
• Google Threat Intelligence (GTI)
• Security Command Center (SCC)
• Various MCP (Model Context Protocol) tools

Contributing

See CONTRIBUTING file for guidelines on contributing to this project.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

---

AI Tool Integration

This repository is designed to work with three AI coding assistants:

Claude Code (claude.ai/code)

• Uses the .claude/ directory for configuration
• Reads context from .claude/rules_bank/ symlink
• See CLAUDE.md for Claude-specific guidance

Cline

• Uses the .clinerules/ directory for configuration
• Reads context from .clinerules/rules_bank/ symlink
• Follows the same runbook and persona system

Gemini CLI

• Uses the .gemini/ directory for configuration
• Reads context from .gemini/rules_bank/ symlink
• See GEMINI.md for Gemini-specific guidance
• Compatible with the same documentation structure

Quick Start for AI Tools

1. Clone the repository:

``bash git clone cd dandye_ai_runbooks ``

1. The symlinks should already be set up. Verify they exist:

``bash ls -la .claude/rules_bank ls -la .clinerules/rules_bank ls -la .gemini/rules_bank ``

1. If symlinks are missing, create them manually:

```bash # For Claude Code mkdir -p .claude ln -s ../rulesbank .claude/rulesbank

# For Cline mkdir -p .clinerules ln -s ../rulesbank .clinerules/rulesbank

# For Gemini CLI mkdir -p .gemini ln -s ../rulesbank .gemini/rulesbank ```

Using with AI Tools

Each AI tool will automatically read the context from its respective directory:

• Claude Code: Reads from .claude/
• Cline: Reads from .clinerules/
• Gemini CLI: Reads from .gemini/

All three tools access the same content through symlinks, ensuring consistency.

Symlink Architecture

The project uses a symlink-based system to share the same content across all three AI tools:

1. Master Content: All documentation, personas, and runbooks are maintained in rules_bank/
2. Symlinks: Each AI tool's configuration directory contains a symlink to rules_bank/
3. Benefits:

• Single source of truth for all content
• Updates automatically propagate to all AI tools
• No duplication or synchronization needed

Configuration Scripts

• set_persona_rules.py: Configure active personas (legacy - may need updates for new structure)
• symlink_common_steps.py: Manage common steps symlinks (legacy - may need updates)
• configure_mcp_tools.py: Configure MCP (Model Context Protocol) tools for Claude Desktop

Example Runbook Invocation Prompts

Use these realistic prompts to invoke various security runbooks. All examples use real data extracted from the reports/ directory:

Alert Triage

Please execute the alert triage runbook for SOAR case 2194.

The case involves repeated authentication failures for user F.K. on host A.S.L from IP 10.1.0.4. Check for duplicate cases and perform initial entity enrichment.

IOC Enrichment

Run the basic IOC enrichment runbook for the suspicious domain scarfponcho.com.

Threat intelligence analysis and SIEM correlation for a domain which was flagged in our environment.

Execute IOC enrichment for the file hash de96a6e69944335375dc1ac238336066889d9ffc7d73628ef4fe1b1b160ab32c found on host WINS-D19.

This was associated with PowerShell obfuscation activity by user LISAWALKER.

Case Investigation

Perform a comprehensive case investigation for SOAR case 3480.

This is a critical priority case involving PowerShell attacks on WINS-D19, potential data exfiltration via BigQuery, and GCP exploitation attempts. Focus on the 17 associated alerts.

Detection Rule Analysis

Analyze detection rule ru_99d1f620-3fe2-41d5-918e-0e1bd2402065 (ursnif_malware_dns).

I need rule validation, tuning recommendations, and analysis of recent detections for this high-severity rule.

Threat Hunting

Execute the IOC threat hunt runbook for IP addresses 46.8.9.205, 46.8.9.206, and 46.8.9.207.

These IPs are associated with the malicious domain scarfponcho.com and need comprehensive SIEM analysis.

Run the advanced threat hunting runbook focusing on lateral movement techniques (PSExec, WMI) around host WINS-D19.

This follows the PowerShell attack activity discovered in case 3480.

Malware Triage

Perform malware triage for the sample with hash de96a6e69944335375dc1ac238336066889d9ffc7d73628ef4fe1b1b160ab32c.

This was involved in the PowerShell obfuscation incident on WINS-D19.

Suspicious Login Analysis

Execute the suspicious login triage runbook for authentication events involving user F.K. on host A.S.L from source IP 10.1.0.4.

This is related to case 2194 with multiple failed followed by successful logins.

Incident Response

Initiate the compromised user account IRP for user LISAWALKER'

Evidence includes PowerShell execution with obfuscation techniques on host WINS-D19 and potential privilege escalation attempts.

Cloud Security Analysis

Run the cloud vulnerability triage runbook for the BigQuery exfiltration alerts in case 3480.

Focus on the DLP context and Google Drive data movement patterns detected by SCC.

Endpoint Isolation

Execute the basic endpoint triage and isolation runbook for host WINS-D19.

This system was compromised in case 3480 with PowerShell attacks and potential malware execution.

Deep Dive Analysis

Perform a deep dive IOC analysis on domain scarfponcho.com using GTI collections.

Include malware family associations, campaign attribution, and infrastructure analysis.

Summary

These examples demonstrate how to invoke runbooks with specific, actionable intelligence that mirrors real security operations scenarios.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: dandye
• Source: dandye/ai-runbooks
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.