I build local-first developer tools for AI agents, security automation, and verifiable workflows.
Current focus: small, auditable CLIs and native tools that make AI-assisted development testable instead of opaque.
- AgentFit tests
AGENTS.md,CLAUDE.md, Cursor rules, Copilot instructions, and coding-agent setup docs like product surface. - Linwarden audits Linux host hardening without root, daemons, network calls, or heavyweight compliance tooling.
- HandoffKit creates deterministic resume packets for interrupted AI coding sessions.
- Flowline is an early macOS local context layer for developer workflows.
- RealityKit Pipeline Guide keeps Blender-to-RealityKit asset work reproducible with CLI gates and evidence.
- RedisInsight maintainers merged an AgentFit-found stale-command fix.
- AgentFit ships as an npm CLI and reusable GitHub Action.
- Linwarden ships on PyPI with Markdown, JSON, SARIF, a composite GitHub Action, and signed releases.
- HandoffKit ships as an npm CLI for Codex, Claude Code, Cursor, Gemini, ChatGPT, and generic-agent handoffs.
TypeScript, Node.js, Python, Swift, SwiftUI, AppKit, React, Next.js, PostgreSQL, GitHub Actions, local-first CLI design, security automation, AI agent workflows.
