360 Secure SkillHub
360 安全技能中心

experiment-framework

该 Skill 提供 Python 运行时与 axexp CLI,用于校验实验规格、规划机器与预算、生成持久化结果包,并在确认提交、机器、仓库、凭据和授权后显式调度远程 ML 实验;执行依赖 Butler 和 RunPod capacity delegate 等外部组件。

antreasantoniou
antreasantoniou
热度 7.65
最近更新 2026-09-12 03:03:22
分类:开发依赖:代码执行 / 读写文件 / 第三方 API应用场景:研究 / 工作流 / DevOps / 自动化脚本

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add experiment-framework
Metadata
name
experiment-framework
description
Validate, plan and explicitly dispatch provenance-linked ML experiments using the bundled Python runtime, with budget reservations and durable result bundles.

Experiment Framework

The runtime lives in this repository's src directory. Install into an isolated Python environment as described in README.md. A nested installable skill is in skills/experiment-framework. Start with axexp --help and axexp validate SPEC --machines MACHINES. Planning and launch without --execute write local bundles but do not dispatch workloads. Never describe this as a no-write preview. Before real execution, confirm the exact source commit, machine, external repositories, credentials, owning budget project and user authorization. Do not weaken the online telemetry, secret handling or atomic reservation/reconciliation checks to make a run succeed. Butler is an external required CLI for execution; its Python entrypoint can be selected by BUTLER_CLI. RunPod execution also requires a compatible runpod-capacity delegate, selected on PATH or by RUNPOD_CAPACITY_CLI. Neither companion is bundled. Missing dependencies are a blocker to execution, not permission to bypass governance. Read README.md for dependencies and limits. Do not treat test-only example specs, a planned artifact graph, or passing unit tests as scientific results or a completed remote run.