Use when you want to make something measurably better and let the agent do the trial-and-error for you — for example "make this faster", "how can I improve this?", "reduce cost/memory", "increase test coverage", or "make this search return better results" (also tuning an LLM training run). You don't
Use when you want to make something measurably better and let the agent do the trial-and-error for you — for example "make this faster", "how can I improve this?", "reduce cost/memory", "increase test coverage", or "make this search return better results" (also tuning an LLM training run). You don't need to know the metric upfront; it helps you set one. It keeps a metric, tries ideas in an isolate
This entry publishes no npm package, so Forge has no dependency tree for it. That is a gap in coverage — not a statement that it has no dependencies.