Catch template JSON errors before Meta's API does

The template-create endpoint is stricter than its documentation reads, and its error messages are famously unhelpful. The most common failure is also the quietest: a mismatch between the variables in your body text and the sample values in example.body_text. Write three placeholders but supply two samples — or forget the example block entirely — and the submission fails with a generic error that says nothing about examples. Developers then burn time second-guessing the body copy, the category, the buttons, everything except the actual cause. Checking the count locally takes a second and removes the guesswork.

Validating before you submit matters for a second reason: review cycles are a real cost. A rejected submission does not just bounce — repeated failed attempts on the same name can slow you down, and if the payload gets far enough to enter content review, you wait on Meta's queue before you can try again. Structural mistakes are entirely mechanical, which means they are entirely preventable. A local lint pass — sequential variables, no placeholder at the start or end of the body, header text within 60 characters, footer free of variables, button set within the allowed combinations — keeps review cycles reserved for what actually needs human judgement: the content itself.

• Name templates like code, not prose. order_shipped_utility_v2 tells a teammate the purpose, category and revision at a glance; template_final_new2 does not.
• Version with a suffix, always. Names are effectively immutable once approved, so a change of copy means a new name. A consistent _v2, _v3 convention keeps old and new versions distinguishable in your dashboard and your code.
• Keep sample values realistic.Reviewers see the filled-in message. “Asha” and “#4821” read like a real notification; “test” and “xxx” invite closer scrutiny.

One caveat worth stating plainly: passing every structural check does not guarantee approval. Meta still reviews the content — category fit, clarity, policy compliance — and no offline tool can predict that outcome. What a linter can do is make sure that when a template is rejected, it is rejected for a reason worth reading, not because a sample value went missing.