agent-skills

Use when running, grading, aggregating, or reporting repository skill evals with skills/skill-eval/scripts/eval_runner.py, when verifying a with_skill/without_skill result before reporting it, or when deciding eval workspace placement, executor/grader separation, model passthrough, or metric capture for an eval run. Do not use for editing the eval suite schema or general skill creation.

Use when making evidence-driven quality decisions for a skill package or its evals from benchmark results, grader feedback, review comments, session-history patterns, trigger failures, or quality regressions; especially when deciding what to change, what not to change, how to update assertions, or whether to rerun skill evals.

Use when building, debugging, porting, or inspecting Minecraft Java Edition mods for Fabric, NeoForge, or Architectury, including Mixins, access wideners, access transformers, mappings, registry/resource issues, NBT payloads, mod JAR inspection, remapping, version migration, dependency API source lookup, GameTest wiring, client screens, HUD rendering, or runtime logs. Also use as a side reference when another planning, execution, or review skill is active and the work depends on Minecraft API facts.