coolbeevip

创建、审阅和优化团队自有项目 README.md，以准确事实和自上而下的信息层级呈现定位、快速开始、使用与维护入口。Create, review, and improve README.md for team-owned projects using verified facts and a progressive top-to-bottom reading flow.

将代码库事实转化为面向业务、产品和管理者的能力说明、场景材料和影响分析。Turn codebase evidence into business-facing capability briefs, scenario narratives, and impact analysis.

从已有代码仓库提取可追溯的功能清单、架构说明和 AI 接手上下文。Extract traceable feature inventory, architecture notes, and AI onboarding context from an existing codebase.

基于 onboarding 产物和源码，围绕功能清单做代码库走读、问答和证据追踪。Guide developers through codebase features with walkthroughs, Q&A, and source-backed explanations.

为已完成的单个 issue 分支推送代码并创建关联 issue 的 GitLab Merge Request。Push a completed issue branch and create an issue-linked GitLab Merge Request.

为已完成的单个 issue 分支推送代码并创建关联 issue 的 GitHub Pull Request。Push a completed issue branch and create an issue-linked GitHub Pull Request.

为具体代码项目维护 Codex 运行时检索层、入口约束、失败记忆和验证 harness。Maintain a Codex runtime retrieval layer, entry constraints, failure memory, and verification harness for a concrete code project.

根据真实反馈、失败案例和执行日志复盘团队技能库，并提出可审核的技能进化建议。Review real feedback, failures, and execution logs to propose auditable improvements to the team skill library.

Unified workflow for MuJoCo modeling and runtime control. Use when the user mentions MuJoCo, MJCF, scene.xml, robot or robot arm scene setup, opening a viewer, actuator control, minimal control experiments, gripper open/close, or model structure inspection. After triggering, route first to MuJoCo Scene Builder or MuJoCo Robot Control through progressive disclosure.