write-setup-authoring-contracts

Provide the accepted task, Candidate, and evidence, then the Skill writes one contract per target under setup-assets/blueprints covering objective, actor, trigger, path, owner, obligations, permissions, and observable outcomes. It first applies the checkout-authoritative writing-for-agents Skill before shaping any contract.

A Setup Authoring Contract is a judgment-only, descriptive specification for exactly one future Rule or Skill target. It records what the target must mean, permit, validate, return, and stop on, while leaving discovery, ordering, tool choice, and generation mechanics to runtime owners.

No. It only authors the blueprint contracts and hands them off as ready input to setup-project-agents, which alone owns downstream generation and setup execution. It never writes future targets, shared SmartKit artifacts, commits, or publishes anything.

The run stops with a structured result: ACCESS_REQUIRED for inaccessible evidence, CONTEXT_REQUIRED for undiscoverable facts, or HUMAN_DECISION_REQUIRED for choices outside accepted authority. Each stop names the owner, consequence, and conditions needed for a fresh run.

Avoid it when no accepted project intent or qualified evidence exists yet, since contracts must be derived from accepted decisions and authoritative sources. It is also unsuitable for implementation recipes, because contracts are descriptive specifications, not procedural instructions.