$ show walkinglabs/learn-harness-engineering
This repository is an educational course on harness engineering for AI coding agents. It is a collection of lecture materials, code examples, and project templates that teach how to build environments, state management, verification, and control mechanisms for AI coding agents. The course is structured as 14 lectures and 8 projects, with a sample Electron knowledge-base app used as a running example. The repository is primarily documentation and educational content, with some reusable templates and scripts for harness creation and auditing.
The LLM identified 5 files worth highlighting for reuse:
A zero-dependency shell script that audits any repository against the five harness subsystems (Instructions, State, Verification, Scope, Session Lifecycle). It checks for AGENTS.md, PROGRESS.md, feature_list.json, Makefile targets, and more, with color-coded output and actionable fix recommendations. This is a reusable tool for evaluating agent harness quality in any repo.
A reusable skill that scaffolds production-grade harness files (AGENTS.md, feature lists, init.sh, verification workflows) for any repository. Includes validation scripts and benchmark tools. Demonstrates a well-structured pattern for generating agent harnesses.
A standalone TypeScript CLI tool that scores a repository on discoverability criteria (AGENTS.md, docs, architecture, feature tracking, handoff files, tests, config, README) and produces a grade (A-F). This is a self-contained, reusable pattern for evaluating repository quality for AI agents.
A TypeScript script that validates the schema and evidence of feature_list.json files, flagging features marked 'pass' without verification evidence. This is a reusable pattern for enforcing feature completion gates in agent-driven development.
A configurable, severity-based file scanner that detects stale artifacts, dead code, structural violations, session cleanliness issues, and configuration problems. It uses a recursive directory walk with depth limits and skip lists, and outputs a categorized report. This is a reusable pattern for maintaining clean repositories.