Agent tools, memory systems, documentation pipelines, prediction-market research, MUD tooling, and small weird interfaces.
I build tools for AI-assisted development: better context, better coordination, richer memory, and sharper command surfaces for agents.
ratacats-skills collects reusable workflows for Claude Code, Codex, Cursor, OpenCode, and other coding agents. Four I reach for most:
Probes a bounded target—codebase, claim, plan, topic, or idea—from many deliberate directions. It closes every probe with evidence or yield, making it useful for hidden-defect hunts, stress tests, and structured brainstorming.
Turns fuzzy repos, plans, diffs, and issue sets into concrete findings, names, acceptance criteria, and next actions. It chains correctness review, plan refinement, module-boundary analysis, cruft detection, and naming review.
Maps concepts and boundaries before proposing the smallest useful rename. It catches overloaded terms, synonym drift, weak contrasts, and names that make code, schemas, plans, or state machines harder to reason about.
Audits whether tests would actually fail under plausible bugs—not merely execute covered lines. It strengthens vacuous, brittle, misleading, or secretly serialized tests without weakening assertions or reducing coverage.
| Project | What it does |
|---|---|
| wideband | Fans one query out across 11 search providers and merges the results into deduplicated sources with provenance and cost telemetry — 4.8× the breadth of any single provider. |
| html-home | Gives repo-local HTML artifacts one localhost front door through manifests, a local index, and a read-only server. |
| pro-cli | ChatGPT Pro and Deep Research in your terminal. |
| raindrop-cli | Exposes Raindrop bookmarks through a clean agent-friendly CLI. |
| discive | Downloads entire Discord channels for local archives and analysis. |
| extra-eyes | Spins up N pair programmers with custom profiles to give instant feedback to one context window. |
| liquid-mail | Shared mailboxes for agents: progress logs, decisions, and coordination context that survive across sessions. |
| knb | Builds knowledge-base workflows for research, synthesis, and durable context. |
| salt-ai | A tiny utility for salting LLM responses. |
| annas-archive-ebooks | Adds ebook search and download workflows for agent-assisted research. |
| ratacats-skills | Reusable, installable agent workflows for evidence-first engineering, research, review, and sharper reasoning. |
| ascii-love | Plays with terminal-native visual interfaces and ASCII display work. |
Villagecraft was a community calendar and organizing platform for neighborhood-scale workshops, skillshares, and hands-on projects.
The work centered on ground-up organizing: helping people coordinate groups, share resources, find spaces, spread the word, and make local co-learning visible. The old FAQ framed it as tools for "hands-on learning activities at the neighborhood level" and a network for co-creation, creative projects, wellness workshops, and community-led making.
| Thread | Repos |
|---|---|
| Agent context and skills | slurp-ai, ratacats-skills, annas-archive-ebooks |
| Agent command surfaces | wideband, pro-cli, raindrop-cli, discive |
| Agent coordination | liquid-mail, pair-programmer-ai, openclaw-hindsight |
| Agent experiments | extra-eyes, knb, ai-isms, dianalokada-skill, gloss, rivet, salt-ai |
| Thread | Repos |
|---|---|
| Browser and visual context | glimpsy, ascii-love, itoa |
| GitHub display experiments | github-marquee |
| Text and social archives | jeffrey-emanuels-tweets |
| Thread | Repos |
|---|---|
| Ranvier and MUD ecosystem | mudcolors, mudcolors-tester, ranvier-community, ranvier-awesome, ranvier-multi-arg-parser, area-respawn |
| Enceladus and game work | enceladusgame, enceladus-client, evermore-color-staging |
| Older web games and exercises | catchphrase, ToEat.ly, bitly_clone, burger_app, json_blob_challenge |
| Thread | Repos |
|---|---|
| Agent and AI systems | ypi, codex-orchestrator, stealth-browser-mcp, LlamaIndexTS, tokens, exa-cli |
| Prediction markets and crypto | pmxt, testnets |
| MUD/Ranvier ecosystem | ranviermud, pinwheel, core, awesome-ranvier, axolemma, neuro, dome-client.js, lobus |
Named without links:
| Area | Projects |
|---|---|
| Prediction-market research and trading systems | pdb, pm, pm-weather, polymarket, trading, econ, pmknb, research |
| Agent memory and knowledge systems | dev-decision-traces, knowledge-scanner, ai-deas, magic-treehouse, tremors |
| MUD automation and game agents | mudbot, mudbot-btree, Our-Mortal-Haul, rune-age-sagas, eterna, ember |
| Browser, media, and research tools | clearfeed, clearfeed-extension, replyhunter, media-scanner, text-crawler, twitter |
| Crypto and finance experiments | sol-fleet, crypto-hourly, xcapit-dist, cryptonation, uniswap-trader |
| Older worldbuilding and web projects | enceladus, enceladus-site, evermore-web, evermore-backup, mudlense, frontman, frontm, puntman, crowd, crowd-symbols, ketascrape |