agent-scaffolding

🛠️ The meta-harness for AI agents — scaffold your own focused, branded agent harness with its own npx CLI, MCP server, memory, learning loop, and witness-signed releases. Works with Claude Code, Codex, pi.dev, Hermes, OpenClaw, and RVM (hardware-isolated sandbox).

A Claude Code skill that scaffolds ready-to-run LangChain Deep Agents and dcode CLI agents into any project.

Scaffold-level recursive self-improvement: does a self-improving scaffold create capability, or only recover engineering slack? Field map + four-claim program, tested on ARC-AGI-3.

Project Horizon Public Repo

Goal-graph scaffolding for LLM agents, in development since 2023. Thirteen pre-registered study series and roughly 600 scored runs measure how recorded advances and refutations reduce loops, improve solving, and transfer proven routes.

The model proposes; the harness disposes. 18+ runnable engineering patterns, framework adapters, and deterministic gates for governed, production AI agents.