Learn Harness Engineering: a project-based course on making AI coding agents reliable

An open-source, project-based course (14 lectures, 8 projects, 15 languages) on building the instructions/state/verification/scope/lifecycle subsystems that make frontier models reliable inside real repos — includes reverse-engineered harness breakdowns of Pi, Claude Code, Codex and DeepSeek, plus loop- and graph-engineering advanced lectures. MIT licensed, ships a harness-creator skill and a zero-dependency audit script.

walkinglabs/learn-harness-engineering19kTypeScriptMIT1 min read

Harness AI Coding Agent Education

SOURCE LINKS

https://github.com/walkinglabs/learn-harness-engineering https://walkinglabs.github.io/learn-harness-engineering/

Related Projects

pbakaus/impeccableApache-2.0

Impeccable: the design language that makes your AI coding agent better at design

An open-source design skill by Paul Bakaus: 1 skill, 24 commands and 61 deterministic anti-pattern detector rules that kill AI frontend slop (Inter everywhere, purple gradients, cards-in-cards). Installs into 16 harnesses including Claude Code, Cursor, Codex and Gemini CLI; live browser iteration plus a Chrome extension. Pure Markdown rules with a zero-LLM detector engine. Apache-2.0, 75k+ stars.

HarnessAI CodingDesign

JavaScript76k4.5k207

OthmanAdi/planning-with-filesMIT

planning-with-files: Persistent file-based planning for AI coding agents — crash-proof markdown plans that survive /clear, compaction and crashes

planning-with-files (27k+ stars, MIT): persistent file-based planning for AI coding agents and long-running tasks — keeps task_plan.md / findings.md / progress.md on disk and lifecycle hooks re-inject the plan into context every turn, so the plan survives /clear, compaction and crashes; a Stop hook forms a deterministic completion gate that holds the agent until the plan reports complete. Project measurements: the on-disk plan cut a 13.3-turn re-orientation to 5.0 turns, 3-of-3 blind A/B wins, 96.7% assertion pass rate. Covers 60+ agents via the Agent Skills standard, with native plugins for Claude Code, Codex CLI, Pi, Hermes, OpenCode and DeepSeek Harness, plus localized skill variants (AR/DE/ES/zh-CN/zh-TW). Pattern follows Manus's "Markdown as working memory on disk"; parallel tasks get isolated .planning/ dirs; /plan-attest locks the plan with SHA-256.

Agent SkillsPlanningHarness

Shell27k2.3k117

HarnessRouter/harnessrouterApache-2.0

HarnessRouter

HarnessRouter Community Edition is the self-hosted, Apache-2.0 implementation of a unified interface for agent harnesses: it turns existing harnesses like Codex, Claude Code, Hermes, PI and DSH into plug-and-play agent backends, so your product runs tasks, retrieves results and switches harnesses through one API. It implements the open Unified Harness Protocol (UHP) and exposes an OpenAI Responses-compatible API, meaning existing Responses SDKs, streaming parsers and UI components work against a UHP server out of the box. The platform handles persistent sessions, streaming progress, files/artifacts, cancellation and structured failures. Inside the container, a three-layer architecture (Console :3000, the only published port, proxies to Gateway :8080 for the Responses API and harness lifecycle, which talks over loopback to Runner :8081 executing harnesses in session workspaces) isolates sessions with separate workspaces and OS users rather than separate containers. Ships with a Console UI, Bring-Your-Own-Key model integrations, custom harness configuration (instructions/tools/skills) and Starter Kits; the UHP spec, machine-readable schemas and a conformance suite live in the repo, with a cloud offering on the same protocol at harnessrouter.ai. For product teams drowning in per-harness integrations, it collapses 36 integration responsibilities into one standardized protocol layer.

HarnessAgent HarnessUHP

Python2.8k29022

CopilotKit/OpenDotsMIT

CopilotKit open-sources OpenDots: self-hostable always-on AI coworkers — each Dot gets its own computer, moving between text, voice calls and Slack, works with any agent harness

CopilotKit open-sourced OpenDots (MIT) on Oct 1: a fully self-hostable template for always-on AI coworkers — its answer to OpenAI's closed Dots (powered by GPT-6 Astra). Each Dot is a specialist agent with a name, role, instructions and tool allowlist, plus its own computer via OpenBot's container supervisor (browser profile and workspace files persist across stop/start). Dots move seamlessly between text chat, realtime voice calls and Slack: calls pair a separate compute agent so long work keeps running mid-conversation, all sharing the same context and tool permissions. AG-UI carries streamed messages, tool calls and agent state between any agent harness and the UI; human-in-the-loop cards pause tool calls for approval; Spaces and Pages hold working documents. Built on TanStack AI + CopilotKit React SDK + Threads + Channels SDK. Clone and customize — enterprise-ready.

HarnessAgent-WorkspaceCopilotKit

TypeScript591733

As an Amazon Associate, we earn from qualifying purchases.