I build small, practical developer tools with a focus on Windows workflows, OAuth callback debugging, Codex troubleshooting, LLM gateway transparency, and AI-assisted diagnostics.

A PowerShell diagnostic tool for Windows URL protocol handlers used by desktop OAuth callbacks.

It helps debug failures where a browser completes OAuth but the desktop app never receives the callback, such as:

codex://oauth_callback?state=...

Highlights:

• Diagnoses HKCU, HKLM, and effective HKCR protocol registrations.
• Detects AppX/MSIX manifest protocol declarations.
• Generates text, JSON, and human-readable explanation reports.
• Includes real-world OAuth callback case studies.
• Maintained with tests, GitHub Actions, releases, and changelog notes.

A Windows diagnostic and repair helper for Codex Computer Use, Chrome, and Browser plugin availability issues.

It focuses on practical failure modes developers hit while using Codex Desktop on Windows, including plugin cache/config drift, missing bundled plugin files, and active-thread tool exposure gaps.

A local security scanner for Codex and OpenAI developer environments on Windows.

It checks for secret-like values, redacts token previews, and flags Codex/OpenAI-looking global npm packages that deserve manual verification.

A CLI tool for auditing OpenAI-compatible LLM gateways and token relay services.

It generates redacted evidence reports for suspicious model mismatch, missing usage metadata, stream inconsistencies, latency anomalies, and gateway transparency gaps without claiming to prove black-box model identity.

• Windows developer tooling
• OAuth and desktop callback flows
• PowerShell diagnostics
• Codex plugin troubleshooting
• Local AI coding environment security
• LLM gateway transparency and model-routing audit trails
• AI-assisted debugging workflows
• Small tools that turn confusing errors into actionable reports

I like tools that come from real problems. OAuth Protocol Doctor started from a real callback failure while connecting a Gmail integration through a desktop app. The project turned that debugging process into a reusable diagnostic workflow.

• PowerShell
• Git and GitHub Actions
• Windows registry and AppX/MSIX diagnostics
• AI-assisted development with Codex

• Building and maintaining oauth-protocol-doctor
• Tracking Windows Codex Computer Use / Chrome plugin failure cases
• Building safer local checks for Codex/OpenAI token and package risk
• Auditing OpenAI-compatible gateways for model and usage transparency
• Documenting real Windows OAuth callback failures
• Improving developer-facing diagnostic output