First rewrite¶

This walkthrough isolates the interception decision itself. It uses the same process-backed lab as First proof, but focuses on what is held, what is edited, and what is actually released.

Capture one complete message¶

ait lab start --exercise delegated-a2a-message --no-open
ait lab trigger
ait intercept pending

Select the pending ID and inspect the immutable original:

ait intercept show MESSAGE_ID

Prepare and release the edit¶

ait intercept edit MESSAGE_ID \
  --set /params/message/metadata/amount=75 \
  --arm attack
ait intercept compare MESSAGE_ID

edit validates the structured body, records the changed JSON Pointer, encodes the complete protocol frame, and forwards it. It does not overwrite the original. The comparison therefore retains both the sender's bytes and the delivered bytes.

Compare the operator decisions¶

After a reset and a new trigger, try the other decisions against a new pending message:

Decision	What crosses the boundary
Original	The message as currently presented, including a live transform if one fired
Sender's version	The immutable sender bytes, bypassing a live transform for this message
Forward modified	The explicitly edited and re-encoded frame
Drop	Nothing
Replay or duplicate	Correlated additional deliveries where the transport supports them

The distinction between Original and Sender's version matters when a transform rule is armed. Confusing them can turn a supposed baseline into a second attack arm.

ait lab stop LAB_ID

Next, use Reading evidence to separate delivery, receiver processing, behavior change, and external effect.