Part

Instructions

You might say: repository guidance, agent context, AGENTS.md

The durable guidance an agent receives about the repository, task boundaries, checks, style, and safety. It may live in AGENTS.md, CLAUDE.md, repository prompts, or equivalent configuration. Keep the root file a short map that points to deeper docs, and write down the "no" decisions so agents stop proposing them again.

What it does

Supplies context the agent cannot infer reliably from the ticket or source tree alone.

Ask

Which instructions are always needed, which are local to a path, and which can the agent verify?

What to look for

1. An AGENTS.md or CLAUDE.md file near the repository root
2. A run log showing loaded instruction files
3. A pull request that reports the checks and conventions it followed

Connections

• See also: Agent harness
• See also: Improvement loop

Sources

1. AGENTS.md, AGENTS.md, Why AGENTS.md?Think of AGENTS.md as a README for agents: a dedicated, predictable place to provide the context and instructions to help AI coding agents work on your project.
2. GitHub, Adding repository custom instructions for GitHub Copilot, IntroductionRepository custom instructions let you provide Copilot with repository-specific guidance and preferences on GitHub.
3. Anthropic, Best practices for Claude Code, Write an effective CLAUDE.mdCLAUDE.md is a special file that Claude reads at the start of every conversation.
4. Kent C. Dodds, kcd-skills, autopilot skillWrite down the "no": decisions agents would otherwise re-propose.
5. Kody, kody-hq/kody repository, docs/contributing/harness-engineering.mdPrefer small, enforceable rules over long, fragile instructions.
6. Matt Pocock, AI Coding Dictionary, AXAGENTS.md, skills, and tools kept lean, so most of the context window is available for the task
7. Matt Pocock, AI Coding Dictionary, AGENTS.mdA file in the environment that the harness loads into the context window at session start — the project's standing brief to the agent.
8. Matt Pocock, AI Coding Dictionary, Progressive disclosureLoading only the context an agent needs right now, with context pointers to the rest.
9. Matt Pocock, AI Coding Dictionary, Context pointerA mention in one document that points to another, so the agent can pull it into the context window only when the task calls for it.

Use it on your project

teardown-apply: hold "Instructions" (instructions) from the teardown at https://teardownatlas.com/software-factory/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal