I design and ship autonomous AI systems — agents that do real work, not demos.
Exploring the frontier of LLMs, multi-agent orchestration and AI-powered infrastructure.
⚰️ container-autopsy — why did the container die? Exit codes, OOM, restart storms → a markdown autopsy report. Zero dependencies. ⬇ 🕯️ death-noticer — an always-on Docker sidecar that writes an obituary for every container that dies, optionally to Slack/Discord. ⬇ ⛪ git-sins — a confession booth for your git history: nine deterministic sins, evidence, penance and a 0–100 Sin Score. ⬇ 🔌 ops-mcp — all of the above as native tools for your AI agent via MCP. The protocol framing is implemented by hand, zero dependencies. ⬇ 🐕 guard-agent — an always-on Strands agent that watches Docker events, diagnoses deaths through ops-mcp and files each incident in your Obsidian vault. AI First, human-approved.
- 🤖 Autonomous Mail Agent — A distributed, event-driven email agent: multi-provider LLM routing, Redis queues, Microsoft Graph, Prometheus + Grafana observability. It reads, reasons and replies to corporate email on its own.
- ✨ repo-glow — Audit your repository presentation and get a Glow Score 0–100. Zero dependencies, shareable badges. Make your repos shine.
- 🛡️ FTG Log Parser — FortiGate firewall logs → clean CSV, for network teams that live in data.
- 🔐 Password Generator — Cryptographically secure password generator, stdlib only (
secretsmodule). - 🏢 Spec Nimbus — An AI sales agency with human-in-the-loop gates: spec-first landings, real demos, autonomous commercial loop.
- AI First: every workflow I touch, I ask "what would an agent do here?"
- Explorer mindset: ship, measure, iterate — the frontier is learned by building.
- Human in the loop: autonomy with control gates. Agents act, humans approve.
- Zero-dependency when possible: stdlib-first engineering.
📫 Reach me: fernedy.arias@gmail.com · 🌐 specnimbus.com
"The future belongs to those who build with AI first, and explore beyond the docs." ⚡