A curated, bilingual (English/Chinese) collection of resources, tools, templates, and implementations for Harness Engineering — the practice of shaping the environment around AI agents so they work reliably.

中文版 →

Harness Engineering sits at the intersection of context management, mechanical enforcement, evaluation, orchestration, and safe autonomy. This repo goes beyond link aggregation: it includes ready-to-use templates, deep concept guides, and a complete map of the emerging ecosystem.

• Foundational Articles
• Concepts & Principles
• Implementations & Loops
• Platforms & Tools
• Templates & Starters
• Skills & Plugins
• Community & Discussion
• Chinese Resources
• Contributing

• Harness Engineering: Harnessing Codex in an Agent-First World — OpenAI's flagship field report. 3→7 engineers, 5 months, ~1M lines of code, ~1,500 PRs. Covers architectural constraints, repo-local instructions, browser validation, and telemetry.
• Effective Harnesses for Long-Running Agents — Anthropic's core guide on initializer agents, feature lists, init.sh, self-verification, and handoff artifacts across many context windows.
• Harness Design for Long-Running Application Development — Anthropic follow-up on task state and evaluator design.
• The Anatomy of an Agent Harness — LangChain's framing: model + prompts + tools + middleware + orchestration + runtime.
• Harness Engineering — Martin Fowler's perspective on the engineering discipline.

Six core principles distilled from the OpenAI field report:

Deep dives in concepts/:

• What is Harness Engineering?
• The Six Principles
• The Ralph Loop Pattern
• From Context Engineering to Harness Engineering

The "Ralph Wiggum Loop" is the canonical Harness Engineering execution pattern: agents run in a loop with fresh context each iteration until the task completes.

Ralph's 6 Tenets mapped to Harness Engineering:

Ready-to-use templates in this repo: templates/

• AGENTS.md.template — Minimal AGENTS.md entry point
• CLAUDE.md.template — CLAUDE.md starter
• PLANNING.md.template — Planning document
• lint/harness-lint.js — Mechanical enforcement example

revfactory/harness: 6 Architectural Patterns

See CONTRIBUTING.md. Prefer resources that are:

• Specific about constraints, evaluation, context management, orchestration, or observability
• Original implementations or primary-source articles
• Useful to practitioners building real harnesses

CC0 1.0