most-capable-agent-system-prompt

A copy-paste system prompt that has your coding agent (Claude Code, Codex, Cursor) build a self-improving agent system for any computer work. Try the lead follo

What is most-capable-agent-system-prompt?

A copy-paste system prompt that has your coding agent (Claude Code, Codex, Cursor) build a self-improving agent system for any computer work. Try the lead follow-up demo at the website link.

Overview of most-capable-agent-system-prompt

What it is

A copy-paste system prompt for coding agents such as Claude Code, Codex, or Cursor. When pasted as the first message, it instructs the agent to build a durable, self-improving agent system for general computer work. The prompt provides a full architecture and operating doctrine including task graphs, verification, file-based memory, evals, governance, and a self-improvement loop. The agent can build either as a harness wrapper around an existing agent or as fresh code built from scratch.

Key features

  • Instructs the agent to inspect the workspace, ask only questions it cannot answer itself, and write a runtime capability matrix and an implementation contract.
  • Sets up a file-based project operating system with files such as AGENTS.md, plan.md, tasks.md, knowledge.md, and FAILURE.md so any compatible agent can continue work from the folder alone.
  • Requires the agent to prove one closed loop end to end as the first milestone: goal, task graph, execution, verification, memory update, visibility to a human, and one lesson learned.
  • Directs the agent to continue building a control plane, workers, memory, model routing, approvals and budgets, an eval harness, and a self-improvement loop in the order specified by the prompt's BUILD ORDER section.

Use cases

Software engineering, scientific research, running a company, data analysis, browser and desktop automation, and complex multi-month projects. The system is designed to learn from every task.

Getting started

Copy the prompt from the code block under The Prompt section of the README. Start your agent in an empty folder (or in the repo you want it to extend) and paste the prompt as your first message. It works with Claude Code, OpenAI Codex, Cursor, Antigravity, OpenClaw, OpenCode, OpenHands, the Claude Agent SDK, or any similar agent that can read and write files and run commands. Answer the few questions it asks, then it starts building.To copy just the prompt from a terminal on macOS:curl -fsSL https://raw.githubusercontent.com/fainir/most-capable-agent-system-prompt/main/README.md | awk '/^text$/{f=1;next} f&&/^$/{exit} f' | LANG=en_US.UTF-8 pbcopyOn Linux, replace | LANG=en_US.UTF-8 pbcopy with | xclip -selection clipboard or | wl-copy. To save it to a file instead, replace it with > prompt.md.

License and pricing

License: MIT. Pricing type: open source (no cost).

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/fainirmost-capable-agent-system-prompt.svg)](https://yunodb.com/tools/fainirmost-capable-agent-system-prompt?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/fainirmost-capable-agent-system-prompt?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/fainirmost-capable-agent-system-prompt.svg" alt="Featured on YunoDB" height="20"></a>