0072 views
Shareable instruction files for Cursor, Claude Code, Copilot, Windsurf, and more — installed alongside your existing agent rules.
**Single source of truth for all agents.** Link-focused Mastodon news reader for iOS/macOS. Pure Swift, SwiftUI + SwiftData, `@Observable` services — no MVVM, no external dependencies. **Linear is the operational source of truth for planning, implementation status, blockers, and verification. Keep it current at all times.**
This guide covers best practices for building environments with `verifiers` and using them to train and evaluate LLMs. It is downloaded automatically using the setup script below (which has likely already been run if you're reading this). See `environments/AGENTS.md` for more details.
Macrograph is a visual programming platform with node-based graph interfaces. Supports both desktop (single-project) and server (multi-project) environments with real-time event processing and type-safe ID management.
This file provides persistent instructions for AI coding agents (Cursor, Copilot, Codex, etc.) working on this codebase. Human contributors should also refer to [CONTRIBUTING.md](CONTRIBUTING.md).
Harbor is a containerized LLM toolkit — a large Docker Compose project with a CLI and a Tauri app for managing AI services. Not to be confused with Harbor container registry which is a completely different unrelated project. This repository, Harbor, is the LLM toolkit.
**Table of Contents** For a broad overview of reinforcement learning, imitation learning and all the training scenarios, methods and options within the ML-Agents Toolkit, see [ML-Agents Toolkit Overview](ML-Agents-Overview.md).
> Create and use specialized AI subagents in Claude Code for task-specific workflows and improved context management. Subagents are specialized AI assistants that handle specific types of tasks. Each subagent runs in its own context window with a custom system prompt, specific tool access, and independent permissions. When Claude encounters a task that matches a subagent's description, it delegates to that subagent, which works independently and returns results. To see the context savings in pra
LLMs excel at **synthesis and critique** when given **bounded, well-structured inputs**. This module wraps provider calls into **role-specific agents** (analysts, researchers, trader, risk) that all speak the same **state dictionary** contract expected by `DecisionExecutor` and `workflow.build_analyst_chain`. Without this layer, prompt strings would be scattered across routes and scripts—untestable and unsafe to evolve.