008 views
Shareable instruction files for Cursor, Claude Code, Copilot, Windsurf, and more — installed alongside your existing agent rules.
The Starling desktop: Swift shell + compositor + apps + the Flutter→Swift framework (`sdk/`), running on the Flutter engine's C core from the sibling repo **starling-engine**, reached through the repo-root `engine` symlink that
**Table of contents**: 1. **User-visible text** — UI copy, toasts, CHANGELOG; never say "订阅" / "subscription" 2. **Hashed domain allowlist is a secret** — never write the pre-image in code, comments, docs, commits, or logs
My name is {YOUR_NAME}. I am a {YOUR_ROLE}, where I {WHAT_YOU_DO}. I enjoy customizing tools and understanding how systems work under the hood. I value open source software for the ability to tinker and modify systems to my preferences.
This file makes you the **orchestrator** for curating the Experimental Factor Ontology (EFO). You read a ticket, plan the work, dispatch three specialist subagents, then commit and open a PR. The goal: **the user assigns a ticket and you take it all the way to a reviewable PR** — creating the branch, doing the work, and writing a clear summary.
MCP server exposing Substack automation to LLM clients. ESM, npm. Development, CI and the image run **Node 24** — `.nvmrc` and `Dockerfile` pin `24.19.0` exactly — while `engines` declares `>=22`, so `src/` may use nothing newer than Node 22 offers.
**No username, no home directory, no local path.** This repo is public. Nothing that goes into it — source, comments, test fixtures, docs, commit messages — may carry the account name or a path off the machine it was written on: not `/Users/<name>/…`, not the
**chimera** is a CLI daemon that manages multiple GitHub Actions self-hosted runners on a single Debian machine. You run it with a config file. It spins up N runner instances, each of which is a fully protocol-compatible drop-in replacement for the official GitHub Actions runner. Each instance connects to GitHub's broker, polls for jobs, executes them, and streams logs back — all concurrently.
Monorepo for the LM Assistant — a self-hosted control plane for Claude Code: session observability and control (web UI + REST API), fleet connectivity over the hub, an MCP server exposing the whole surface inside Claude sessions, missions/automation, cross-host memory, and service connectors.
Instructions for Claude Code (and other AI assistants following the same convention) working in this repo. Think of this as project-specific context that should load before you read any other file. If you are a human reader: this file tells the assistant how we want it to behave on Foveacast. You can skim it to see what the expectations are. The long-form equivalents for humans are [CONTRIBUTING.md](CONTRIBUTING.md) (how we work), [docs/PRD.md](docs/PRD.md) (what we're building), and [LEARNINGS.
> **You are running inside a self-paced workshop.** This repository is a > bootcamp that *you*, the agent, guide the student through — one module at a > time — recording progress as you go so later sessions can pick up where the
> Read this file at the start of EVERY session before touching code. > These are the load-bearing rules. Breaking one silently is how this project > rots across sessions. If you must break one, it’s a deliberate, isolated
A GPU-rendered (GPUI, à la Zed/Ghostty), Rust, Excel-compatible spreadsheet built to be **stupid-fast on huge sheets** (Excel-max = 1,048,576 rows × 16,384 cols). Engine = **IronCalc**; UI = **GPUI** (custom raw-gpui grid + gpui-component for chrome).