run-provenance
对可复现研究至关重要,便于审计与重现实验流程。
记录运行环境、输入与随机性,确保每次结果可追溯重现。
把一次运行的“是什么产生了这个结果”彻底记录下来:运行环境(解释器、锁定的包、系统层等)、输入字节(数据集、配置文件)和随机种子都写入 meta.json。在要保存实验结果、两台机器结果不一致或要验证“数据在两次运行间是否变化”时用。亮点是把环境、输入和种子三条可追溯链都钉死,不靠记忆,便于精确复现。
▸ 展开 SKILL.md 英文原文
Use when a kept run must be attributable to exactly what produced it — the environment it ran under (interpreter, locked packages, domain engines, system layer, native architecture), the input bytes it consumed (datasets, external files, config/constants), and the random draws it made (seeds recorded and re-derivable). Everything `results/<name>_<date>/meta.json` must record about a run, and how to make each field true. Also use before any campaign whose results will be kept, when two machines disagree on a result, and when asking "did the data change between these two runs". Trigger phrases: "pin the environment", "lockfile", "which versions produced this", "works on my machine", "environment drift", "hash the inputs", "did the data change", "which dataset version", "fingerprint the config", "data provenance", "record the seeds", "which seed produced this", "unseeded randomness", "make this run reproducible", "what produced this number".
帮我安装这个 skill:https://raw.githubusercontent.com/huguryildiz/ALETHEIA/main/skills/core/run-provenance/SKILL.mdcurl -fsSL "https://raw.githubusercontent.com/huguryildiz/ALETHEIA/main/skills/core/run-provenance/SKILL.md"# run-provenance
"What produced this number?" must have one answer, computed and recorded, not remembered.
Every kept run carries a meta file — `{{evidence_dir}}/<name>_<date>/meta.json` — and that
file must pin three provenance chains, each of which dies silently if not captured at run
time: the **environment** it ran under, the **inputs** it consumed, and the **seeds** it
drew from. "It worked last month" is not provenance — it is a memory.
## When to use
- Before any campaign that lands in `{{evidence_dir}}` (preflight) — the meta file needs all
three chains.
- Creating or onboarding a project without a lockfile; adding a dependency or upgrading an
interpreter/domain engine.
- Laun