cybersec-mcp
MCP server with 323 cybersecurity prompts and 7 chained workflows. Install it and Claude (or any MCP-compatible client) can run an incident-response plan, a cloud audit, or a pentest by calling tools instead of you copy-pasting prompts.
Live demo ยท MIT License ยท Model Context Protocol
Install
npx -y @xu-c0/cybersec-mcp
Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"cybersec": {
"command": "npx",
"args": ["-y", "@xu-c0/cybersec-mcp"]
}
}
}
Then:
Use cybersec to plan an incident response for unusual outbound traffic from a SIEM-flagged host. SIEM is Splunk, EDR is CrowdStrike.
The agent picks the incident-response scenario, fills your variables in, and walks through detection, triage, containment, eradication, and post-mortem with concrete commands at each step.
323 prompts across 8 categories. Every prompt takes typed variables and returns output in a defined shape (steps, tables, SIEM queries, MITRE tags).
| Category | Prompts | Covers |
|---|
| Red Team | 45 | Pentest methodology, AD attack paths, C2 infra, social engineering |
| Blue Team | 42 | Log analysis, IR playbooks, detection engineering, deception |
| SOC Operations | 42 | Splunk/Sentinel/Elastic queries, alert triage, runbooks, shift handover |
| Cloud Security | 38 | AWS/Azure/GCP audits, IAM, container security, CSPM |
| OSINT | 38 | Domain intel, threat actor profiling, footprinting, attribution |
| GRC | 38 | ISO 27001, SOC 2, NIST CSF, risk assessment, policy generation |
| Vulnerability Analysis | 42 | CVE triage, CVSS 4.0, patch prioritization, pentest reports |
| AI Agent Security | 38 | LLM red teaming, prompt injection, agent guardrails, supply chain |
Source: content/prompts-master.md โ generated web-app/js/data.js.
Scenarios
Seven end-to-end workflows that chain prompts and pass variables between steps:
- Web App Penetration Test โ recon โ mapping โ fingerprinting โ API testing โ exploitation โ post-exploit โ reporting
- Incident Response โ detection โ log investigation โ severity โ containment โ eradication โ comms โ lessons learned
- Cloud Security Audit โ IAM โ network โ storage โ database โ logging โ compliance
- Bug Bounty Recon โ subdomains โ ports โ tech โ OSINT โ surface โ vuln assessment
- Compliance Audit (ISO 27001) โ scoping โ risk โ controls โ evidence โ gaps โ docs
- Threat Hunting โ hypothesis โ query design โ pivot โ validation โ response
- AI Security Assessment โ inventory โ access control โ red team โ prompt injection โ supply chain โ monitoring
Definitions live in web-app/js/scenarios.js.
Web demo
cybersec-mcp.vercel.app โ browse every prompt, fill in variables, copy the rendered text into any LLM. Dark mode, English / ํ๊ตญ์ด / ๆฅๆฌ่ช, no signup. Same data as the MCP server, different interface.
Useful when you want to inspect what a tool will send before wiring up the server, or hand a teammate a one-off prompt.
ATT&CK mapping
Red team, blue team, and SOC prompts are tagged to MITRE ATT&CK tactics. The full mapping is in ATTACK_MATRIX.md โ useful for purple-team exercises and detection-coverage reviews.
Layout
mcp/ MCP server (TypeScript, in progress)
web-app/ Static demo deployed to Vercel
content/ prompts-master.md โ prompt source of truth
examples/ Client configs (Claude Desktop, Cursor, Claude Code)
parse_prompts.py regenerates web-app/js/data.js from content/prompts-master.md.
Contributing
PRs welcome โ new prompts, MITRE tags, scenario workflows, translations, MCP tool fixes. Schema and quality bar in CONTRIBUTING.md.
This project is for authorized security testing, defensive operations, security research, and education. PRs promoting unauthorized access will be rejected. See CODE_OF_CONDUCT.md.
License
MIT โ see LICENSE. MITRE ATT&CKยฎ is a registered trademark of The MITRE Corporation; this project is not affiliated with MITRE.