Skip to content
View JacobStephens2's full-sized avatar

Block or report JacobStephens2

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
JacobStephens2/README.md

Jacob Stephens

Safe AI automation for revenue-critical legacy systems. I design, build, and operate production systems end to end — a multi-portal PHP / MySQL reservations platform, multi-tenant AI assistants on Docker, and a Python agent-orchestration host, for a multi-million-dollar specialty-travel business. That's the Legacy-to-AI Bridge problem: modernization without a rewrite, agents without blast radius, and human-in-the-loop where it matters.

Most of that work lives in private repositories. The architecture and the reasoning behind it are written up in the ETA Platform case study and in infrastructure-patterns (sanitized ADRs, a threat model for agent sandboxes, and an operational review checklist).

Shipped apps

Products I've built and shipped end to end — offline-first, client-side encrypted, and released to real users across web, iOS/watchOS, and Android.

AppWhat it isLiveCode
Chart35Privacy-first Creighton Model fertility charting — offline-first PWA, end-to-end encrypted sync, provider sharing; in the App Store / Play / TestFlight, with users via organic searchchart35.com/creightonchart35-showcase (public architecture & privacy slice)
DromeFree, local-first iOS tracker for 75-day challenges — one Rust rules engine driving native SwiftUI + WidgetKit, event-sourceddrome.dayshowcase coming

Both apps run on the same pattern I use in infra work: one core, native shells, and privacy by construction. Server-side code stays private; the public showcase demonstrates the front end and the architecture.

In the works

Apps I'm building toward a public release; the code stays private for now, but each has a public home you can follow.

AppWhat it isWhere to follow
QuadrilleLifting log — a native iOS + watchOS weightlifting tracker with a shared core and integrated Watch app, built to ship to the App Storequadrille.app
NightlochPartner-locked digital accountability curfews — WireGuard + AdGuard Home DNS filtering with a Go control planenightloch.app
ClaveCooperative salsa partner-dance card game for two players, with a headless engine and two-device roomsclave.dance

Production tools

Command-line and self-hosted tools I built for my own agent + infra workflow, in daily production use.

ToolWhat it isCode
vaulted-agent-launcher (va)Launches Claude Code, Codex, and Grok with vault-resolved secrets in-process — per-agent blast radius, no .env files or vault tokens on disk (1Password, Bitwarden SM, pass, sops; macOS + Linux)vaulted-agent-launcher
muxboardFlask-embeddable web dashboard over tmux for one host or a fleet, live in-browser attach — default-deny auth, attach caps, documented threat modelmuxboard.dev · code
skillsReusable SKILL.md agent skills for application planning, adversarial review, and adjudication — used across my own Claude Code / Codex workflowsskills
inkvokeSingle-binary, agent-friendly Go CLI for OpenAI's image models (gpt-image-2 by default) — generate from prompts, edit existing images, and run batches from a JSON manifestinkvoke.dev · code

Published packages

Small, dependency-free primitives extracted from that production work — each ships as a wire-compatible TypeScript + Go pair, tested against parity vectors so the two produce byte-identical output.

PackageTypeScript (npm)Go (pkg.go.dev)
HMAC webhook verifiers (GitHub / Stripe / Twilio / Mandrill)@stephenspage/webhook-verifywebhook-verify-go
AES-256-GCM envelope + PBKDF2 vault + zero-knowledge sealed shares@stephenspage/webcrypto-envelopewebcrypto-envelope-go

TypeScript releases are published via GitHub Actions with npm Trusted Publishing (OIDC) — no long-lived tokens, provenance statements on every version.

Upstream contributions

More landing in the Model Context Protocol ecosystem — that's where the "safe AI automation" thesis externally validates.

Available for platform-infra roles.

📄 Resume · Portfolio · Available for Staff / Lead platform-infra roles


St. Carlo Acutis, pray for us.

Pinned Loading

  1. infrastructure-patternsinfrastructure-patternsPublic

    Sanitized ADRs and infrastructure patterns from production systems I have designed and operated

  2. webhook-verifywebhook-verifyPublic

    Tiny typed HMAC webhook verifiers - raw-body (GitHub-style), timestamped with replay tolerance (Stripe-style), and URL+sorted-params (Twilio, Mandrill) - constant-time compares, zero dependencies.

    TypeScript

  3. muxboardmuxboardPublic

    Flask-embeddable web dashboard for managing tmux sessions across one host or a fleet, with live in-browser attach. Default-deny auth, attach caps, and a thorough threat model.

    Python

  4. webcrypto-envelopewebcrypto-envelopePublic

    Tiny typed Web Crypto wrapper: PBKDF2 keys + AES-256-GCM envelope, plus seal/open for zero-knowledge shareable payloads

    TypeScript

  5. cascadecascadePublic

    Waterfall white-noise player: one headless Rust core driving six native shells — web (PWA), Android, macOS, Windows, iOS, watchOS. A headless-core + native-UI kata.

    Swift

  6. k3s-demok3s-demoPublic

    Kubernetes/k3s demo: stateless service + production-grade manifests (probes, limits, hardened securityContext, HPA, ingress, kustomize). Statically validated with kubeconform.

    HTML