Define the structured compaction output contract

What problem would this solve?

Compaction providers differ in how they return reasoning, structured final output and completion status. A clear shared contract would make summary handling and fallback behavior consistent across providers.

What would a good outcome look like?

Agree on when structured output may replace a plain-text compaction summary, including incomplete responses, and preserve useful fallback behavior in both agent loops.

Possible approaches

Should structured compaction require direct leading structured output, or retain support for reasoning-wrapped output and carry trusted completion status through the provider and compaction interfaces?

The decision should specify supported response shapes, behavior when completion status is unavailable, and compatibility expectations. Verification should cover complete/incomplete responses, ordinary plain-text fallback and both agent loops.

Additional context

This is a design question; no implementation is proposed before the issue reaches Ready.

• I have verified this does not duplicate an existing feature request

Do not begin implementation until the issue reaches Ready on the Goose Issues board.