write-quality-contract

Guardrails for an agent OS are defined by drafting a quality contract that codifies a non-negotiable rule, verification checks, anti-patterns, banned vocabulary, and tone to ensure consistent behavior and safer outputs.

A quality contract for agent governance is a specification that defines verification checks, anti-patterns, and tone to enforce consistent style and safety. You need one after domain discovery and identity model setup to ensure auditable governance across all skills.

Verification checks and banned vocabulary are set up by creating a domain-input/quality-contract.md file. This specification structures source verification, freshness checks, and audience alignment alongside a catalog of common failure patterns with concrete fixes.

Yes, a quality contract explicitly handles empty context and external data sources by defining guidelines with explicit trust and verification rules, ensuring the agent OS safely processes missing context and validates external data.

A limitation of this quality contract approach is that it requires explicit mappings to domain-input artifacts and a clearly defined zero-setup entry point, meaning it cannot function effectively without prior domain discovery and identity model setup.

No, the agent OS quality contract requires no specific dependencies or components to function. It operates as a standalone specification that bootstraps OS governance through the domain-input/quality-contract.md file.