I build small developer tools and reference architectures around explicit contracts, reviewable decisions, and careful automation. My current work sits where AI-assisted development meets platform engineering: provider boundaries, safe Git and file workflows, repository QA, and cloud-system design.
| Project | What it demonstrates |
|---|---|
| AWS Public Change Alerting | An explainable AWS reference architecture with numbered specifications, ADRs, JSON Schema contracts, and deterministic validation. |
| CliSpecForge | A compact Python reference implementation for turning Markdown CLI specs into reviewable file sets through one model request and guarded writes. |
| Personal Agent Skills | Reusable Codex and Claude Code workflows for repository verification, Git publishing, PR checks, branch cleanup, and README work. |
| Compact Theme | A framework-free HTML, CSS, and JavaScript theme used across my GitHub Pages sites. |
- Tight, inspectable Python tools with zero-dependency cores and optional provider adapters.
- Agent workflows that preserve user authority around Git, files, reviews, and external systems.
- Architecture artifacts that can be validated before a runtime exists.
- Honest evidence: tests and fixtures prove their measured path, and docs name the remaining gaps.
Claude OKF repo kit is archived. Its useful repository-governance ideas now inform the smaller tools and skills above. spec-drift is also being retired after its experiment concluded.
AI agents · Developer tooling · Platform engineering · Specs & ADRs · Cloud systems