Bench Setup — Git skill for Claude Code
Build a new benchmark scope tailored to the user's task.
How to install Bench Setup
Installs to ~/.claude/skills/ant-open-skills-custom-model-bench-bench-setup/SKILL.md
mkdir -p ~/.claude/skills/ant-open-skills-custom-model-bench-bench-setup && curl -fsSL https://raw.githubusercontent.com/ant-open-skills/custom-model-bench/HEAD/commands/bench-setup.md -o ~/.claude/skills/ant-open-skills-custom-model-bench-bench-setup/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Bench Setup does
description: Build a new benchmark scope tailored to the user's task. Four-question intake plus a provider picker; accepts a GitHub repository URL.
Build a new benchmark scope from scratch, tailored to what the user is actually working on. Take the user through a deliberate intake — friction-with-attraction, not a template wizard. The questions themselves should signal that the kit is taking their task seriously.
All file operations below write into `${CLAUDE_PLUGIN_ROOT}/skills/custom
Alternatives in Git
- Perfup — Autonomous performance optimization: research, PoC, benchmark, implement, review, PR 3.6k ★
- Skill Copilot Provider — GitHub Copilot CLI as optional zero-cost provider via copilot -p programmatic mode 2.8k ★
- Claude Code Python Feature List & PR Roadmap — 面向社区贡献者的能力清单、路线图与 PR 指南 231 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
BDD From Code
Backward-engineer BDD from existing source code. Accepts a local path OR a git repo URL (with optional branch)
Factory Review Signoff
Supervisor-triggered review sign-off for a factory campaign branch — lock (collapsed campaign scope), gather t
Forge Report
File a harness failure as a Forge diagnostics ticket (Firestore → GitHub issue, no GitHub account needed). Cal
GitHub Activity Agent
Local GitHub activity monitor with zero dependencies. A dashboard shows pushes, PRs and reviews across your re
Agent Intern
MCP bridge that drives four coding CLIs — Google's Antigravity (agy), OpenAI Codex, GitHub Copilot, and Cursor
Git Agent
Coordinate issue intake, branch work, and merge steps.
Related Agents
Bench Reporter
Updates benchmark documentation with results from bench-runner. Maintains BENCHMARK.md, README.md, and EVAL-GA
Benchmark Sweep
Runs the real benchmark/performance-gate tasks defined in a repo's own deno.json(c) (today, @zanix/server's be
Resume Writer
Generates tailored resumes as .docx files. Use when given a job posting (URL or text) to create a resume match