Trigger Quality Decides the New Batch Too

We have written before that the description line decides whether a skill ever fires. The new batch is another test of that claim.

Trigger quality is the first dimension in the rubric and the one most people treat as documentation. It is not documentation. It is the routing layer: the sentence an agent reads to decide whether this skill is relevant to the request in front of it.

Across the forty new skills, trigger quality averages 5.96 with a standard deviation of 2.66. Seven skills score below 3.0. The correlation between trigger quality and the final score is r = 0.933 (mechanical, because trigger is weighted at 20%, but the direction is clear).

The explicit-trigger group

Eight of the forty files contain an explicit Use when clause. Those eight average 8.28 overall and 8.31 on trigger quality. The other thirty-two average 6.42 and 5.37.

This is an observational split, not a controlled experiment. The eight skills that write explicit triggers are also more likely to be longer and more carefully written. But the size of the gap is hard to ignore.

The eight are: k12-lesson-plan-creation, crm-migration-consolidation, cs-operations, neon-rutger-lead-routing, pipeline-visibility, zendesk, dynamic-pricing-ecommerce, and product-description-generator.

The counterexamples

cover-letter has a trigger score of 1.5 and a content score of 8.5. The content is useful; the file just does not say when it should be used. An agent scanning dozens of skills has no reason to select it.

At the other end, dynamic-pricing-ecommerce scores 9.1 overall with a trigger score of 9.0. It starts with the user's question — "what should I charge for this product?" — and builds the procedure around that question.

The practical fix is cheap: name the situations, the adjacent requests that should not trigger the skill, and the exact phrases a user is likely to say. The rest of the file can stay the same.

For the original argument, see the one line that decides whether your skill ever fires.