Onboard Agents

Guide the user through a conversational interview that produces a complete, project-specific AGENTS.md or CLAUDE.md for how an AI coding agent should behave in this repository.

Separation of Concerns

This skill produces the behavior layer of the agent instruction stack.

Keep content here only when it directly improves recurring agent behavior. If a detail is better described as project background, stack fact, architecture explanation, process handbook material, or nearby reference documentation, use a canonical companion document and link to it instead of restating it here. These layers must not duplicate each other.

Canonical companion documents may include openspec/config.yml or openspec/config.yaml for project DNA, ARCHITECTURE.md for system structure, CONSTRAINTS.md for externally imposed boundaries, EPISTEMIC-MAP.md for validated facts vs open questions and assumptions, and JARGON.md for internal terminology.

Use the rule below to decide what stays in this file and what belongs in a canonical companion document.

Hard Rule: What Does NOT Belong Here

If material does not answer "how should the agent behave?", it does not belong in AGENTS.md or CLAUDE.md.

Move that material to the appropriate canonical companion document when one exists, or leave it out.