Agentic engineering · DevSecOps · Knowledge transfer

I build multi-agent systems on large-scale legacy code, at the intersection of applied AI and security.

PHP developer who came to the trade through a career change — ~7 years in, now working deep inside a multi-million-line legacy codebase (~50,000 files). Day to day I push two things forward: agentic engineering (multi-agent workflows, MCP servers, context engineering) and application security as a contributor to my company's security team. I also spend a fair amount of energy on the part most people skip: making AI legible to non-experts.

Building          → agent ecosystems, MCP servers, security tooling
Working terrain   → PHP legacy at scale + DevSecOps
Direction         → Applied AI engineering with security as the field
Currently leveling → English, measured RAG in production, Security+

• 🔐 mcp-noctua — MCP server exposing a sandboxed pentest toolbox (sqlmap, nuclei, ffuf, httpx, naabu, katana…) to an LLM orchestrator, for authorized security audits. Agentic security, offensive side.
• 🔌 mcp-obsidian — filesystem-based MCP server for an Obsidian vault: read/search/write notes directly on disk, no REST plugin required.
• 🔌 mcp-redmineApi — MCP server driving a Redmine instance from Claude via its REST API (issues, time logging, metadata, allow-listed uploads).
• 📐 claude-md-playbook — a field-tested playbook for writing CLAUDE.md files: treat them as infrastructure — triggered rules, testable claims, converged through use (EN + FR).
• 🧠 claude-obsidian-system — a personal knowledge architecture: an Obsidian vault as a compounding second brain (LLM-Wiki + Zettelkasten + Evergreen) wired to Claude Code skills, MCP and persistent memory.
• 🛠️ cc-switcher — switch Claude Code authentication between multiple accounts (personal / work) with a single shared ~/.claude.
• 🧰 DevRewind — local, single-user activity journal to prep annual reviews: PHP/SQLite REST API (prepared statements, FTS5 search, PHPUnit-tested) + Vue 3 / TypeScript frontend.
• 🎲 rpg-soundboard — self-hosted touch soundboard for tabletop RPG sessions: recursive sound library, per-tile overlays, Range-streamed audio, PWA/offline, optional local Kokoro TTS. Plain PHP + vanilla JS, no build.

Flagship work lives inside private/legacy repos: a role-separated sub-agent ecosystem (reviewer, security sentinel, devil's advocate, pre-mortem) and a Semgrep security skill bundle, both operating on a ~50,000-file PHP codebase.

AI / Agents — Claude Code · MCP · context engineering · multi-agent orchestration · local inference (Ollama) Security — Semgrep · SAST · DevSecOps (Jenkins/Snyk) · system hardening Linux — Kubuntu / KDE daily driver