trellis-update-spec

To document code contracts after debugging, you capture executable signatures, payload fields, and error behavior into .trellis/spec documents. This ensures future AI agents or developers can reproduce safe behavior across different sessions without relying on chat history.

Full executable spec depth is mandatory for command, API, or database contract changes, migrations, and cross-layer infrastructure integration. It records concrete validation matrices and boundary error behavior so future work can execute safely.

Standardizing spec updates for AI workflows requires a consistent seven-section template covering scope, signatures, contracts, validation matrices, cases, tests, and wrong versus correct examples. This keeps executable code specifications separate from thinking guides.

Yes, you can use it to record validation and error matrices during refactoring. It captures good, base, and bad cases alongside required test points, ensuring the refactored code maintains safe behavior and conforms to the original executable contracts.

Yes, the spec format supports capturing wrong versus correct examples when relevant triggers are met. This clarifies boundary conditions and prevents future implementation mistakes by explicitly showing incorrect approaches alongside the correct executable code.

Executable code contracts are needed for cross-layer infrastructure to standardize request and response payloads, environment keys, and error handling. Without them, integration knowledge is lost to chat history, making safe reproduction across sessions impossible.