Use Codex to Review SendGrid Delivery Symptoms From a Test Export
Classify an example delivery problem and identify the evidence needed to distinguish delivery delays from permanent rejections.
中文Setup for Codex
Use Codex and synthetic aggregate event counts. A production analysis would require authorized provider evidence, comparable dates and the appropriate sender domain.
- Open the pinned Twilio AI repository revision and copy this skill directory, with its companion files, into your Codex project's .agents/skills directory.
- Open Codex in a scratch project and select the matching Twilio SendGrid skill. Use the local test fixture before configuring credentials for any live API check.
Example task
Provide a mock monitoring sample with 80 delivered, 15 deferred and 5 bounced out of 100 distinct messages during a fixed one-hour window. Ask the SendGrid advisor to compute the rates and list the evidence needed to determine why the deferred messages have not completed.Expected result
The local check yields 80% delivered, 15% deferred and 5% bounced at the snapshot. The report asks for later event outcomes and sanitized diagnostic reasons before offering a cause.
Upstream documentation
Original SKILL.md on GitHub · MIT license · Source reviewed
Frequently asked question
Can SendGrid event counts alone reveal why an ISP deferred delivery?
Counts establish the size and timing of a symptom. Diagnostic reason fields, sender context and later message outcomes are needed to narrow the cause.