基于 SOC 职业分类
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/UCL-ERL/skills --skill example-skill命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Audits benchmark design, baselines, metrics, ablations, statistical reporting, leakage risk, and claim-result alignment for Robotics and AI papers. Use when the user asks whether experiments are convincing, fair, reproducible, or sufficient for paper claims.
Builds a portable, source-grounded Markdown dossier from a remote research project, including method state, experiment design, result ledgers, figures, failures, provenance, and paper-ready claim candidates. Use on the machine that contains the code and experiment outputs when preparing a handoff for local paper writing, collaborator review, or project archiving.
Audits whether a Robotics or AI manuscript accurately describes its implementation, configs, data pipeline, training and inference behavior, evaluation code, defaults, and generated tables or figures. Use when reconciling a paper with a codebase, preparing an artifact release, checking implementation drift, or verifying that equations, algorithms, hyperparameters, and system claims match the released code.
| name | example-skill |
| description | Use this skill when an agent needs to perform one concrete, reusable workflow. |
Use this skill when:
Do not use this skill when:
Run or describe checks that prove the workflow worked.
# Example:
python scripts/validate_catalog.py
State the artifact or condition that proves the skill is done. Avoid vague criteria such as "make it useful."
Keep this section short. Add links to primary documentation or papers if the skill depends on external knowledge.