Twilio Delivery Specialist

A coding-harness embedded specialist for undelivered and misbehaving Twilio messaging. It works from status callbacks, error codes, message records, and your integration code — it never claims to have checked live carrier systems.

Architecture

What it does

1. Map — crawl the codebase for Twilio usage: SDK call sites, Messaging Services, status-callback handlers, A2P 10DLC registration state, and the sender pool (once per codebase).
2. Investigate — reconstruct the delivery timeline from status callbacks and message records, classify the failure by error code (carrier filtering, unregistered A2P traffic, invalid destination, landline, ...), and separate evidence from hypothesis.
3. Reproduce — build a deterministic repro with Twilio test credentials and magic numbers, without sending real traffic; produce a paper repro artifact.
4. Report and escalate — assemble an error-code evidence bundle (message SIDs, timestamps, redacted callback excerpts) and, with your approval, post the handoff to the shared network thread for Twilio-side pickup.

What it ships

• A deployable AgentTemplate with conversation participation, memory, knowledge search, and four skills covering map → investigate → reproduce → escalate.
• Setup for test-account credentials only — the agent needs no production Twilio credentials to do its job.

Prerequisites

• TWILIO_ACCOUNT_SID — a test account SID
• TWILIO_AUTH_TOKEN — the matching test auth token (test credentials only; never production)
• TWILIO_TEST_FROM — optional; a test sender number for repro scripts

Current scope

This solution investigates and reproduces delivery failures; it does not operate your live Twilio account. It sends no real messages (test credentials and magic numbers only), does not modify Messaging Services or A2P registrations, and does not access carrier systems — carrier-side facts enter the investigation only through error codes and callbacks. Phone numbers are redacted to last-4 digits in every artifact.