benchmark-assistant
Benchmark Assistant
Purpose
benchmark-assistant helps the user actually use the benchmark runner.
It is a meta skill. It does not replace scripts/run-benchmark.js, and it does not define domain quality by itself. It chooses useful benchmark commands, runs or prepares them, explains results, and hands failure diagnosis to skill-evaluator when needed.
When to Use
Use this skill when the user says things like:
More from huajiexiewenfeng/thinking-skills
content-creator
Use when the user is developing articles, essays, posts, newsletters, scripts, talks, titles, outlines, arguments, audience positioning, drafts, or content structure.
2learning-coach
Use when the user wants to understand a concept, learn a topic, build a mental model, explain something in simpler terms, find knowledge gaps, make a study plan, practice retrieval, or turn confusing material into usable understanding.
2thinking-router
Use at the start of a user request to classify intent and route to the most appropriate domain-specific thinking skill. Do not assume software development unless explicitly indicated.
2skill-evaluator
Use when reviewing a Thinking Skills failure, golden case candidate, user feedback about a skill response, eval result, routing mistake, unsafe behavior, overly long output, too many questions, jargon exposure, regression risk, or when deciding how to improve or preserve a skill.
2technical-deep-dive
Use when the user needs technical analysis involving code, repositories, architecture, debugging, performance, APIs, systems, databases, implementation trade-offs, tests, deployment, or source-level reasoning.
2conversation-review
Use when the user asks for self-review, Dolores mode, conversation review, skill trace audit, failure analysis, eval gap detection, improvement-loop suggestions, failure case or golden case status, or skill feedback dashboard.
2