skill-creator

You create OpenCode skills by defining scope and success criteria, then generating SKILL.md frontmatter and executable instructions, which are iteratively tested through evals and benchmarks to ensure trigger accuracy.

Eval-driven skill development automates prompt generation, eval execution, and feedback collection across iterations to optimize skill triggering and performance under real workloads.

Yes, skill-creator enforces safety controls, dependency tracking, and organized resources including scripts, references, and assets to enable reproducible deployment of OpenCode skills.

The best way to structure an OpenCode skill is through an intake interview to capture intent, followed by drafting the initial SKILL.md and a basic eval set for iterative optimization.

Yes, skill-creator supports benchmark and eval iterations to validate performance and trigger accuracy, ensuring the skill meets real workload demands before deployment.

Iterating through evals is necessary to validate trigger accuracy and optimize performance, ensuring the skill performs reliably under real workloads before production deployment.