AI Lab · LabChan · Tracker Row 13

UC1b Loop-Guard Correction

Correction of the shared Discord TX-bot authorization problem discovered during UC1b attempt #1, followed by a successful end-to-end live retest.
UC1b Attempt #2 — PASS

Result

The shared LabChan TX bot can now carry an explicitly authorized LabChan From: participant through the redesigned RX loop guard. The new live test was accepted, archived once, and invoked CC1 exactly once.

Commit8898f1c
Discord message1545086439350337547
Automated tests63 passed in 0.83s
RX statusaccepted
ArchiveExactly 1 entry
InvocationCC1 exactly once

Why the correction was needed

UC1b attempt #1 successfully transmitted through the shared Discord TX bot, but RX rejected the message because the bot's Discord author identity was not a normal 1:1 participant identity. The correction adds an explicit, default-deny authorization path for approved shared transport identities without weakening the existing participant identity checks.

Implemented correction

  1. contract/registry.md gained an Authorized Transport Identities section under explicit-change governance.
  2. The shared LabChan TX bot's Discord transport identity was authorized initially to send on behalf of the human operator.
  3. RX now distinguishes a registered participant identity from an approved shared transport identity.
  4. A shared transport is accepted only when its LabChan From: value appears on that transport's allow-list.
  5. Unauthorized originators receive loop_guard_unauthorized_transport_originator; unknown identities remain loop_guard_unmapped.
  6. The authorization table is parsed by the listener and supplied to the RX processor at startup.

Validation

CC1 added coverage for the authorization-table parser and for authorized, unauthorized, unknown, and no-authorization transport cases. The full suite completed with 63 passed in 0.83s.

CC1 then stopped the old listener, verified that no stale listener remained, restarted RX with the corrected code, confirmed a single listener instance, reviewed the scoped diff, committed and pushed the change, and ran a fresh UC1b attempt.

Live UC1b Attempt #2

The fresh TX message was posted by the shared TX bot, matched the new authorization entry, and passed the RX loop guard. The Discord message was accepted; the archive contained exactly one entry, CC1 was invoked exactly once, and the watermark advanced.

Conclusion: the TX → Discord → corrected RX → CC1 path worked end to end. Attempt #1 remains preserved as FAIL evidence; attempt #2 is the successful validation.

Repository and evidence

The correction was pushed as commit 8898f1c, with evidence and a full report preserved in the lab's private operations repository.