opencode-effect-enforcer: More than 30 Effect skills plus a write-time pattern matcher that catches places where Effect offers a cleaner solution. I want my codebases gilled up on Effect. It gives me confidence that the code is correct and keeps agents decisively in the Effect latent space, rather than treating it as an afterthought
opencode-imagen: So that I can say "use imagen to make a sick backsplash for this page that fits the aesthetic based on what you see in @docs/vibes/" and it actually happens. Uses gpt-image-2 and tokens are billed not separately but through the hooked-up monthly Codex subscription
effect-oxlint: Wraps the oxlint plugin SDK with Effect so it's more ergonomic to write all your custom oxlint rules Effect-firstly. Kinda pointless but I do use it and some other people do too
oxlint-plugin-effect: Custom lint rules for my projects, mostly the lint-rulified versions of the patterns enforced at write-time by opencode-effect-enforcer
scrape-docs: OpenCode v2 has a feature where you can add a reference to a public GitHub repo and OpenCode then maintains a clone of it locally so I can @reference it in any chat and the agent is able to navigate it with bash/filesystem tools. So this repo is a scraper that scrapes the following documentation sites every 5 minutes and maintains a separate repo for each of them, effectively ensuring I am always able to e.g. @foldkit-docs in an OpenCode session and trust that I am pointing the agent to the most up-to-date version of that documentation as a locally-cached collection of markdown files: Alchemy, Better Auth, Drizzle, Electron, Ent (Go), Ent (TypeScript), Foldkit, LadybugDB, Pi, PlanetScale, ts-morph, Turborepo
starred: Much like the list you are reading, this is a categorized and easier-to-skim list of every repo I have starred on GitHub
workspace: Mostly all my dotconfigs
yazi: Fork of the original yazi because I use it so much and had some bespoke customizations I wanted to make
mpsuesser (meta!): To get a profile header like this, one must create an eponymous repo and put things in the README
interview-me-phonecall: I had an 8-hour drive the next day, wanted to talk out some ideas and be sure I could use the transcripts after
pi-effect-harness: This became opencode-effect-enforcer when I switched from Pi to OpenCode around the time they released the v2 beta
pi-compound: Had too many things going on when this was incepted so I did not spend the time to ingratiate it and thus it never stuck
effect-autoagent: Low effort few-prompt experiment in applying the then-novel effect harness to reproduce the then-novel original autoagent project but with Effect. In hindsight this was a terrible fit and serves merely as a cautionary tale
effect-claudecode: I do not use Claude Code because perhaps to my detriment I am keenly lock-in averse but when Fable was released I brought this up to date with Claude Code's plugin spec at the time but then da gubament took it away so I never really used it. Maybe some day
oxlint-plugin-alchemy: Too much stuff, and moving too fast, and also not the greatest fit for a lint plugin to begin with
oxlint-plugin-foldkit: Devin released an official Foldkit plugin and to my delight harvested whichever rules from this repo were worthy so now I just use that
effect-prodigi: Good binding for a Print-on-Demand SaaS app, Prodigi seemed like the best vendor to me at the time this was made, but my excitement for the project was alas fleeting and so this seemingly complete repo is here but its original intended consumer remains private/unfinished
opencode-sketch: I think this worked but I quickly thereafter moved my primary sketching surface to Dojo
wxt-github-readme: Tiny web extension to add copy-to-clipboard buttons to the front-pages of GitHub repos for copying the README, so that if I just want to copy the README I don't need to spend an extra 5 seconds / 1 page load / 2 clicks. I forget why I stopped using it
ReposCLI: This was important to me for a few hours but then I realized I do not have sufficient runway to allocate my time to such things
For about three years my obsession was indie game development, and because I am hardcore but also because these are the kinds of games I grew up on I only ever worked on multiplayer client-server games.
RUN Football was to WaLK Football what DotA 2 was to DotA.
Spire Two was the WoW Challenge Mode experience crystallized into a game that would never ask you to level up a character or to procure the optimal set of gear to qualify for participation; all about the dungeon run.
Line Tower Wars 2 was to Line Tower Wars what DotA 2 was to DotA.
USP-Ninjas-archive: This was like my first big project, wrote this when I was 14, it's a plugin for Counter Strike 1.6. There was idk like 100 maybe 200 people who would regularly play this game USP Ninjas on a custom map that was made specifically for it. But it was just a map, and there were rules to the game, so sometimes people would break the rules and the only way to referee was to start yelling about it in all chat. And you had to manually facilitate like picking teams, restarting the rounds, keeping track of the real score, and so on. This plugin formalized and enforced all of that stuff.
wc3fball-discord-lobby-bot-archive: We had a small community of people who loved to play WaLK Football, which requires 8 people to play. It's a Warcraft 3 custom game so you have to launch the game, log in, join the lobby, and wait for 8 people, but there's all these subtle frictions around actually needing people to sit and wait with WC3 open and idle in the game lobby for 30+ minutes. So this was a Discord bot where people could !ready and !unready in Discord, and when the ready count got to 8 it would ping everybody so we could all fire up the game and join the lobby and go.