I build Python and JavaScript tools for the parts that fail quietly: webhooks, plugin discovery, and release handoffs.

My favorite demo is the one that breaks on purpose.

A signed-webhook simulation with SQLite-backed duplicate/conflict detection and bounded retries. Event identity survives restarts; exhausted retries require explicit recovery instead of quietly trying again.

Python · SQLite · HMAC · unittest — independent sample; synthetic events, simulated CRM, no customer traffic.

A zero-dependency CLI and GitHub Action that verifies skills and hooks against a pinned Codex release. Declared capabilities and observed discovery stay separate; installation alone is not proof that a plugin works.

JavaScript · Node.js · GitHub Actions · Docker — experimental; strict isolation is Linux-only, and discovery is not a security certification.

A release-review interface with an executable acceptance model. Missing evidence keeps the fictional release blocked; unit and browser tests cover the decision rules, state transitions, and keyboard-focus recovery.

HTML/CSS · JavaScript · Playwright — independent frontend sample, not a deployment controller.

In python-docx-ng PR #131, I fixed quote-sensitive style lookups by binding XPath variables and adding regression coverage. The maintainer reviewed the edge cases and merged it. One open-source contribution, not a client case study.

I'm extending the webhook sample into HandoffLab: a failure-replay matrix that checks downstream side effects and records the operator's recovery decision. The local prototype is tested; it is not released yet.

For a focused software project or open-source collaboration: email me or find me on LinkedIn.