菜单
Use when migrating a PySpark job or function in this repo (jobs/) to Polars (fargate/), or when asked to "migrate to polars", "convert this PySpark function", "port this job to fargate". Walks the repo's migration checklist so nothing from the PR template gets missed.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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/NMDSdevopsServiceAdm/DataEngineering --skill polars-migration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。
Use before/when investigating an OOM or memory blowup in a Polars pipeline job, or comparing candidate fixes/implementations by memory usage, or when asked to "add diagnostics", "instrument this job for memory", "profile this pipeline run". Documents how to run the `RunDiagnostics` utility (`polars_utils/run_diagnostics.py`) via a throwaway Fargate task, choose a sampling interval, and read back its S3 output.
Use when starting a new piece of work — "/new-ticket", "start a new ticket", "let's start on ticket 1234". Runs branch check → short interview → hands off to Plan Mode for the SPEC. Pairs with the `commit-push` and `open-pr` skills for the rest of the cycle.
Use when a ticket's work is ready for review — "/open-pr", "open a PR", "raise the PR". Run once, at the end of a ticket. Pairs with `new-ticket` and `commit-push`.
基于 SOC 职业分类
正在显示 SKILL.md
| name | polars-migration |
| description | Use when migrating a PySpark job or function in this repo (jobs/) to Polars (fargate/), or when asked to "migrate to polars", "convert this PySpark function", "port this job to fargate". Walks the repo's migration checklist so nothing from the PR template gets missed. |
This repo is mid-migration from PySpark (jobs/) to Polars (fargate/). When migrating, follow this checklist — it mirrors .github/PULL_REQUEST_TEMPLATE/polars_migration_template.md.
utils/ → project utils/ → repo-wide projects/utils/. Only add a new util if nothing suitable exists at any of those scopes.| Used by | Location |
|---|---|
| One job/dataset only | projects/<project>/<stage>/utils/*.py |
| Multiple jobs/datasets in a project | projects/<project>/utils/*.py |
| Multiple projects | projects/utils/utils.py |
LazyFrame for test data, not eager DataFrame.pytest-pattern skill for the repo's dataclass+parametrize convention..collect() happens where it does).
# converted to polars -> path/to/new_fargate_file.py
CHANGELOG.md — see CLAUDE.md's "Changelog" section for the format.Scope discipline: only migrate what's being asked. Don't mass-rewrite untouched PySpark files just because you're nearby, and don't refactor beyond what the migration needs (see the repo's general no-premature-abstraction guidance).
For scale/memory concerns specific to the new Polars code (laziness, .over() vs joins, streaming-engine coverage), see CLAUDE.md directly, or the polars-streaming-check / over-vs-join skills if the migrated logic touches those areas.