Freelance web developer

GitBasedDocs

A private docs site that reads Markdown from one GitHub repo. Write in Obsidian, push, and a few seconds later the pages are live for the people you shared them with. Each project is a folder in the repo, and a reader sees only the projects an admin linked to their account. Anyone else gets the same 404 as a project that does not exist. A webhook starts a delta sync that fetches only changed files, a timed recheck covers a missing webhook, and pages are served from SQLite, so reading never calls GitHub. It renders the Markdown Obsidian and GitHub both know: Shiki code, Mermaid diagrams, KaTeX math and Obsidian callouts, plus wikilinks and image embeds. Next.js 16, React 19, next-auth v4 and Drizzle on SQLite, about 9,200 lines of TypeScript, MIT.

• Tailwindcss
• Typescript
• Nextjs
• ShadCN
• React
• Bun

MD Kanban

A kanban board where every card is a markdown file on your disk, built so you can watch what an AI coding agent is doing without asking it. Each connected folder is a project, each subfolder groups work, and each task is one `.md` file whose YAML frontmatter holds status, priority, assignee, due date and labels. Dragging a card between columns rewrites the file, and the agent edits the same files from its side. That is the whole design: Jira and Linear keep state behind an API, so an agent needs credentials and a connector before it can move a card, while a folder is something it can already write to. The repo ships the contract twice, as `AGENTS.md` for any agent and as a Claude skill, and tracks its own remaining build as task files inside itself. Vue 3, Vite, TypeScript and Tailwind v4 over the browser File System Access API, about 1,800 lines, MIT.

• Tailwindcss
• Javascript
• Typescript
• Vue

AI Writing Rules

A Claude Code plugin that stops Claude writing like an AI. It loads a 270-word ruleset into every session, then runs a checker after every file write that flags the tells by line number and refuses to let the model move on. The rules come from 29 documented patterns, each researched into its own file with a measurable threshold: burstiness under 0.4 where human prose sits at 0.6 to 1.2, more than 20 em dashes per 1,000 words where humans use 3.2 to 10, more than one polished three-item list per 200 words. The checker is 17 regex rules and 3 statistical ones in Python with no dependencies. Tested against a sample with 14 planted tells, it caught all 14, stayed silent on clean prose, and found a real tricolon in the research files themselves.

• Python