OpenCode Skill Creator brings Anthropic-style evals

Systematic evaluation is the critical "missing link" for transitioning local LLM agents from experimental toys to reliable production tools. The tool automates the creation of "should-trigger" and "should-not-trigger" test sets, replacing manual trial-and-error with empirical benchmarks. By decoupling Anthropic's methodology from Claude-only environments, it supports over 300 models with iterative optimization loops and integrated visual reporting for human-in-the-loop verification.