Skill Posting Optimizer

Problem

When publishing a model skill, the issue is often not missing functionality, but an unclear posting surface: names look like internal drafts, descriptions mix implementation details, invocation notes, and user value; field naming is inconsistent, and name, desc, and example text use different wording, making it hard to see the problem it solves, the expected input, and the likely output.

How It Works

The skill is oriented toward cleaning up skill posting copy. Given a draft, it can check consistency across title, summary, field descriptions, and example wording: whether the title points to a concrete object, whether the description covers input, output, and intended use case, whether field names are duplicated, vague, or conflict with the body, and whether the wording reads more like engineering documentation than marketing. The output is typically revision suggestions and uncertainty flags that need human confirmation.

Boundaries and Notes

It fits pre-release copy review and wording alignment, not source review, permission checks, security audits, or runtime validation. Final capability must be grounded in the actual implementation and reproducible behavior; optimized copy should not be treated as an API semantic, stability guarantee, or permission boundary.