Hello, I'm
Javier
Morales
AI agent builder and AI security engineer.
AI Security Engineer | Agentic AI & MCP Security | Securing autonomous agents in production | Detection & Response
About
AI agent builder and AI security engineer. I design autonomous AI agents for security operations, and the guardrails, evaluations and governance that keep them safe and accountable.
Architect and operator of a 24/7 agentic AI system that runs security operations across a large multi-account AWS environment. LLM agents triage and investigate alerts, call tools across SIEM, identity, cloud and collaboration platforms through the Model Context Protocol (MCP), and escalate to a human analyst only when it matters. One design principle holds: the model proposes, deterministic code decides, and a human approves anything irreversible.
Specialist in securing the AI itself: layered prompt injection defence, screening of tool outputs before they reach the model again, sandboxed agents with deny lists, and signed single-use approvals in a multi-agent orchestrator. Trust in AI is measured, not assumed: every model change passes an evaluation gate, accuracy is monitored with automatic rollback, and agent memory is protected against poisoning.
Author of 'Securing Autonomous AI: Patterns for Controlling Agents, Models & Tools' and creator of open-source AI security tools: a linter for MCP servers and AI agents, guardrails for AI coding assistants, and a prompt injection detector. 15+ years across infrastructure, cloud and security. MSc in Cybersecurity and co-author of a peer-reviewed benchmark of Android malware analysis tools (MDPI Electronics, 2024). AWS Security and Machine Learning certified.
Key Impact
Autonomous Security Operations
Production AI-driven system integrating SIEM, identity, cloud APIs, ticketing, and vulnerability management through MCP architecture. Purpose-built agents handle tiered investigation workflows from automated triage to deep behavioral analysis.
Detection-as-Code at Scale
Hundreds of SIEM detections across multi-account AWS with unit testing, contextual enrichment, dynamic suppression, and automated deployment. Led a systematic noise-reduction program across the alert pipeline.
AI Agent Defensive Architecture
Multi-layered defensive framework for AI-assisted operations: integrity verification, prompt injection scanning, supply chain monitoring, and guardrails enforcement against emerging AI agent attack classes.
What I Build
Agent Audit Kit
Self-hosted security linter for AI agents and MCP servers. Point it at an mcp.json or agent config and it flags risks mapped to the OWASP Agentic Top 10: excessive agency, prompt injection, tool poisoning, supply chain, SSRF, data exfiltration and missing human-in-the-loop controls. Outputs JSON, SARIF for code scanning and an HTML report, and ships Sigma detection rules to monitor the same risks in production. Pure Python, zero dependencies, no telemetry. Tested against 64 public MCP server repositories.
View project →AI Security Guardrails
Open-source defence system for AI coding assistants such as Claude Code. Pre and post tool-call hooks stop risky actions before they run: prompt injection detection across 14 pattern categories with Unicode evasion handling, credential leak scanning, dangerous-command and privilege-escalation blocking, and SHA256 integrity verification. Pure Python, 68 automated tests, MIT licensed.
View project →Prompt Injection Detector
Lightweight Python library for detecting prompt injection attacks in AI agent outputs. 23 regex patterns, NFKC Unicode normalisation, zero-width character stripping, and credential leak scanning.
View project →MCP Security Toolkit
Model Context Protocol servers for security operations: SIEM querying, threat intel enrichment, and incident response automation. Connects LLMs directly to security tooling for autonomous SOC workflows.
View project →Detection-as-Code Framework
Production detection rules with unit testing, contextual enrichment, and dynamic suppression for multi-account AWS environments. Includes real-world examples of SIEM detections and systematic false-positive reduction.
View project →n8n Security Workflows
Ready-to-import n8n workflow templates for security operations: AI-powered alert triage, threat intel enrichment, certificate expiry monitoring, daily security digests, and new user auditing.
View project →MotoRadar — Price Intelligence for Motorcycles
Side project: full-stack price comparison engine for second-hand enduro and trail motorcycles in Spain. Aggregates listings from 4 platforms, applies z-score deal scoring, tracks price trends, and detects outliers. Built with FastAPI, PostgreSQL, HTMX, and a custom browser extension for data collection.
View project →Technical Stack
AI & Automation
Detection & Response
Cloud & Infrastructure
Security & Compliance
Backend & Data
Frontend & Scraping
Experience
Staff Security Engineer
Tealium · Remote, Canary Islands
- Architect and operator of a 24/7 agentic AI triage system in production: the model proposes, deterministic code decides, and a human approves anything irreversible.
- 87% of high-priority alerts resolved automatically at about 97% precision, with a circuit breaker that reduces autonomy if accuracy drops.
- Agentic AI security by design: layered prompt injection defence, tool-output screening, agent memory hardened against poisoning, and an LLM evaluation gate in CI.
- Multi-agent orchestrator in shadow deployment: sandboxed agents, HMAC-signed approvals, autonomy tiers and a kill switch.
Senior Security Engineer
Tealium · Remote, Canary Islands
- Built the detection-as-code platform in Panther SIEM: hundreds of detections across cloud, identity, SaaS and endpoint, with unit tests and automated deployment.
- First AI-assisted triage workflows and the MCP integration architecture connecting AI agents to security tools.
- Compliance-as-code for SOC 2, ISO 27001 and NIST 800-53, plus security monitoring and SIEM coverage for HIPAA-regulated environments.
- Event-driven re-architecture from EC2 to AWS Lambda with S3 triggers; SIEM data-lake queries optimised from over 3 minutes to around 20 seconds.
Cyber Security Engineer
The Workshop · Malaga, Spain
- Zero Trust across AWS and hybrid environments: IAM hardening, service control policies, WAF and centralised audit logging.
- Operated NGFW, IDS/IPS, WAF, SIEM and EDR; continuous compliance (SOC 2, ISO 27001, NIST) with centralised logging.
Platform Engineer
The Workshop · Malaga, Spain
- Hybrid cloud on OpenStack and VMware with Infrastructure as Code (Terraform, Ansible AWX, Puppet) and CI/CD pipelines.
- Multi-site infrastructure: network segmentation, hardening and disaster recovery.
Senior Cloud Engineer
eSentire · Cork, Ireland
- Managed AWS infrastructure for managed detection and response (MDR) services, ensuring high availability and security posture
- L2/L3 incident management for VMware vSphere/vROps. Designed ETL pipelines and optimised database queries for high-volume data processing
- Technical lead on infrastructure projects, overseeing planning, execution, and SLA delivery
L2/L3 Support Engineer
VMware · Cork, Ireland
- Expert support for VMware vSphere and vROps across enterprise environments
- Led multi-platform troubleshooting across networking, storage, and virtualization
Earlier Career
Havas Media, Wincor Nixdorf, Roche, A&O Systems + Services · Ireland & Spain
- Havas Media, Wincor Nixdorf and Roche (Madrid, 2013-2016): IT systems and operations, VMware administration, DBA (PostgreSQL/MySQL/Oracle) and service desk L1-L2.
- A&O Systems + Services (Las Palmas, Jul 2010 - Nov 2013): IT support engineer on client placements at Xerox (printer fleet diagnosis and repair) and Vodafone (L1-L2 IT support and service desk).
Education & Certifications
MSc Cybersecurity
UNIR — Universidad Internacional de La Rioja · 2019 — 2021Co-author of a peer-reviewed benchmark of Android malware analysis tools (MDPI Electronics, 2024). DOI: 10.3390/electronics13112103
Higher Diploma in Cloud Computing
National College of Ireland · 2017 — 2019Diploma Multiplatform App Development
IES El RinconDiploma Telecommunications Systems
IES PolitecnicoCertifications
Books & Resources
Practical guides and free tools for securing AI systems.
Securing Autonomous AI
The practitioner's guide to defending AI agents in production — threat models, defense patterns, and incident playbooks.
Get the book → FreeDetection Engineering Starter Kit
Free: 3 production-ready detection templates and AI triage prompts for security operations.
Download free →Latest from the Blog
264,000 People Are About to Install an AI Agent Harness. Run This 5-Minute Audit First.
Everything Claude Code crossed 264k stars and is genuinely well built. That popularity is the point: popular is not the same as audited by you. The 5-minute audit I run against any agent harness before I trust it.
Read more →I Ran My Agent Auditor Against the Top Agent Repos. It Found Real Bugs, and Embarrassed Me.
I ran my open-source agent auditor against two of the most-talked-about agent repos on GitHub. It found real unpinned MCP servers, out-of-model hooks and risky installers, and it also flagged injection defenses as attacks and nouns as destructive actions. The honest before and after.
Read more →Your Blocklist Is Matching Inside Words
I lost real data for weeks to a three-letter blocklist term matching inside an innocent word. Exclusion filters don’t error when they’re wrong — they quietly remove things, and ‘zero results’ looks identical to ’no such data.’ A story about substring matching, word boundaries, and why the most dangerous filter is the one you never see fire.
Read more →