菜单
Compare multiple agent memory systems by analyzing their code, benchmarks, and architecture
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。
基于 SOC 职业分类
Stage and commit current changes with an auto-generated conventional-commits message. From main, first creates a feature branch. Does NOT push.
Push unpushed commits on the current branch to GitHub and open a PR via gh CLI. Aborts if there are no unpushed commits. Ignores working tree state entirely — for uncommitted changes use /commit first.
Search arxiv for academic papers on a given topic, return structured summaries
Deep-dive into a GitHub repo - analyze architecture, code patterns, and design decisions
Conduct a structured research survey on a topic, combining academic papers and open-source projects
| name | bench-compare |
| description | Compare multiple agent memory systems by analyzing their code, benchmarks, and architecture |
| user_invocable | true |
| arguments | [{"name":"projects","description":"Comma-separated list of projects to compare (e.g. mem0,zep,letta)","required":true},{"name":"criteria","description":"Comparison criteria (e.g. performance,api-design,scalability)","required":false}] |
Compare multiple agent memory projects across key dimensions.
${projects}:
${criteria} if provided):