Problem
Some workflows compose multiple side effects across boundaries, for example:
- payment
- database write
- notification
Each side-effect boundary should own its own execution guard, but retries still need a shared correlation mechanism.
Proposal
Add support for a parent execution_id or correlation ID so:
- each boundary keeps separate execution ownership
- composed workflows still converge safely under retries
- tooling can inspect related execution records together
Why this matters
This allows Execution Guard to support:
- cross-tool retries
- payment + DB write workflows
- multi-step agent actions
- partial failure recovery
Problem
Some workflows compose multiple side effects across boundaries, for example:
Each side-effect boundary should own its own execution guard, but retries still need a shared correlation mechanism.
Proposal
Add support for a parent
execution_idor correlation ID so:Why this matters
This allows Execution Guard to support: