Recipient identity changed after approval
Recipient-binding gate should fail before send.
SAMPLE-002 applies the BitEvo method to one outbound message. The critical boundary is not merely tool access; it is the exact recipient, approved content, execution window and confirmation path.
“Can send messages” is not decision-grade authority. The ledger narrows the effect to one recipient, one approved revision and one allowed execution path.
Send one approved outreach message to one resolved staging/test recipient.
Exactly one recipient identity bound to one approved contact record.
Synthetic Outreach Operations owner for SAMPLE-002.
Draft approved message → one send attempt to the approved recipient within the allowed window.
No bulk send, recipient substitution, attachment upload, production campaign launch, reply impersonation or contact creation.
No automatic resend after ambiguous delivery state; reconcile provider state before another send.
Synthetic operator decides reconcile / cancel / repair / retest.
Evidence Before Effect requires the recipient identity and message revision to remain bound to the authorization at execution — not merely when the draft was created.
The send target must resolve to the same approved contact identity used during authorization.
Exact message body/template revision must match the approval record.
Recipient and approval state must be revalidated at send time; any changed identity or revoked approval blocks the effect.
The example permits the send only inside the synthetic approved window attached to the decision trace.
Template/config revision must match the approved revision.
Provider event or message record must bind the send attempt to the exact recipient and message revision.
CONSTRAIN: stop before send, or after ambiguous acknowledgement block resend until provider state is reconciled.
These scenarios are illustrative and unexecuted. A real engagement would define provider-specific replay and data boundaries in written Rules of Engagement.
Recipient-binding gate should fail before send.
Content-approval gate should fail.
Authority should collapse before effect.
Treat as potential False Green; do not infer completed send.
Retry should remain constrained pending reconciliation.
Authority/time gate should fail.
Route to review; do not choose silently.
Treat external effect as untrusted.
The finding is intentionally a worked format, not a reproduced event. It demonstrates why tool permission alone is too coarse for outbound authority.
SAMPLE-AUTH-002
Authority mismatch / recipient-binding gap
SYNTHETIC WORKED EXAMPLE — NOT EXECUTED
The send path retains a valid tool permission while the recipient identity or approved message revision no longer matches the decision evidence.
One outbound send to one approved recipient using one approved message revision.
Not observed. SAMPLE-002 does not claim a real send or provider response.
A generic send permission cannot justify an effect when recipient/content binding has diverged.
CONSTRAIN send authority whenever recipient, approval, message revision or provider confirmation is ambiguous.
Should this workflow keep autonomous send authority when recipient/content binding can drift between approval and execution?
CONSTRAIN
The Authority Budget is broader than the evidence contract if send permission survives identity or approval drift.
Bind authorization to recipient identity + exact message revision + send window; require recipient-bound provider confirmation before retry.
The workflow may prepare a send, but execution stops when any binding evidence is missing or changed.
Actual provider idempotency, delivery semantics and event latency are NOT TESTED in this synthetic example.
Replay drift and ambiguous-ack paths; require stop/review or exact recipient-bound confirmation before any resend.
A repair is decision-relevant only if drift and ambiguous delivery paths now end in a bounded state.
If any binding is unresolved, the workflow must stop or route to review rather than send or resend.
THIS SAMPLENOT TESTEDCompare this pack with CRM writes and deployment/config changes to see which controls stay invariant and which become effect-specific.