A collection of independent developer utilities, AI-assisted applications, and experiments built around a simple principle:

Solve actual problems. Question assumptions. Keep humans in control.

Portable agent skill and CLI for keeping requirements, decisions, and project documentation aligned with the code and development status.

Focus: Documentation integrity · Agent portability · Validation

Three standalone tools designed to work independently:

• Checkpoint Orchestrator: Session-isolated task checkpoints, progress recovery, and SQLite-backed persistence.
• Git That Shit: Recoverable snapshots of important development configuration and agent state before risky changes.
• Stranger Danger: Detection, redaction, and auditing of sensitive information at AI-agent boundaries.

A Docker-based environment for reproducible plugin tests, compatibility checks, offline simulations, and failure isolation. Test the plugin before the plugin tests your patience.

AI-assisted publishing workflows for Bludit CMS, including writing, editing, SEO audits, and model-provider configuration. Existing content changes stay subject to human review.

A local-first MCP server connecting AI clients to Bludit CMS for content retrieval, search, and controlled edits with a propose → preview → apply → verify workflow.

A personal-training management application in development, designed around practical workflows: client records, workout programming, session logging, progress tracking, and printable reports.

An HR decision-support project focused on retrieving policies, grounding answers in sources, and offering practical guidance.

Developer experience: TypeScript · JavaScript · Python · PHP · SQL · CLI tools

Applications and infrastructure: React · Node.js · Express · PostgreSQL · SQLite · Linux · Docker · GitHub Actions · Cloudflare

AI workflows: OpenCode · MCP · Portable Agent Skills · Local models · OpenAI-compatible APIs

Problem spaces: Automation · Privacy · Human-centered AI · HR technology · Process improvement · Documentation quality

• Useful beats elaborate. Build for a real problem, not for a fashionable stack.
• Evidence over assumptions. Verify behavior through tests and reproducible results.
• Independent by design. Avoid unnecessary lock-in and fragile dependencies.
• Human oversight matters. Consequential AI-assisted changes deserve review.
• Documentation should match reality. Not optimistic fiction.
• Don't ask why. Ask why not.

Build something useful. Break assumptions. Fix what's broken. Repeat.