Agent Skills › ymx10086/ResearchClaw › experiment-tracker

experiment-tracker

GitHub

用于在ResearchClaw中轻量级记录实验参数、结果和笔记，支持列出历史运行及对比不同运行的元数据或指标。适用于用户需要追踪实验详情或比较多次运行差异的场景。

src/researchclaw/agents/skills/experiment_tracker/SKILL.md ymx10086/ResearchClaw

触发场景

用户想要记录实验参数或结果用户需要查看或过滤历史实验运行用户明确要求对比多个实验运行的差异

安装

npx skills add ymx10086/ResearchClaw --skill experiment-tracker -g -y

更多选项

非标准路径

npx skills add https://github.com/ymx10086/ResearchClaw/tree/main/src/researchclaw/agents/skills/experiment_tracker -g -y

不安装直接使用

npx skills use ymx10086/ResearchClaw@experiment-tracker

指定 Agent (Claude Code)

npx skills add ymx10086/ResearchClaw --skill experiment-tracker -a claude-code -g -y

安装 repo 全部 skill

npx skills add ymx10086/ResearchClaw --all -g -y

预览 repo 内 skill

npx skills add ymx10086/ResearchClaw --list

SKILL.md

Frontmatter

{
    "name": "experiment-tracker",
    "emoji": "🧪",
    "triggers": [
        "experiment",
        "ablation",
        "metrics",
        "compare runs"
    ],
    "description": "Log experiments, list past runs, and compare experiment metadata or metrics. Use when the user wants lightweight experiment tracking inside ResearchClaw."
}

Experiment Tracking

Use this skill when the user wants to record experiment parameters, results, notes, or compare multiple runs.

Tools

log_experiment: append a structured experiment record
list_experiments: filter and inspect previous runs
compare_experiments: compare selected runs side by side

Guidance

Normalize names, tags, and metric keys when the user provides loose wording.
Use log_experiment after the user shares a concrete run or result.
Use compare_experiments when the user explicitly asks for differences across runs.