mma-context-blocks

You can register large documents once as context blocks and reference them by ID in subsequent mma-* calls. This avoids re-sending content, minimizing data transfer across iterative audit, debug, and plan workflows.

Sharing specs and logs across iterative tasks is done by storing them once as immutable context blocks. You then reference the generated ID in later mma-* calls to reuse the same content without re-uploading.

Context block lifecycle management applies server-side storage with immutability, TTL lifecycle, and deletion semantics. This ensures safe reuse by keeping the registered document content stable until its TTL expires or it is explicitly deleted.

Yes, you can reference the same document ID across different audit and plan workflows. Registering the document once assigns a reusable ID that any subsequent mma-* call can reference to retrieve the shared content.

No, you do not need to re-upload your spec file for new plan tasks. Once the spec is registered as a context block, the system applies server-side storage so you can reference its ID across all future tasks without re-transmitting the content.