staff-level-elixir

Use this skill when writing, reviewing, or refactoring Elixir, OTP, Ecto, or Phoenix/LiveView code. It corrects the staff-level judgment calls a capable model gets wrong by default — reaching for a GenServer as the universal tool, rescuing instead of letting processes crash, N+1 Ecto access, non-atomic writes, unbounded/linked Task fan-out, atom exhaustion from user input, eager Enum where Stream fits, and calling Repo from the web layer. Applies whenever the work touches process design, error handling, concurrency, data access, or LiveView, even if the user doesn't name them.

跳到安装

仓库: pproenca/dot-skills
最近来源活动: 2026年7月5日 21:48
检测到的 SKILL.md 语言: 英语
星标: 211
分支: 18

安装方式

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。

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/pproenca/dot-skills --skill staff-level-elixir

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。

检查来源文件

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。

文件资源管理器

28 个文件

SKILL.md

来源说明 · 只读预览

name: staff-level-elixir
description: Use this skill when writing, reviewing, or refactoring Elixir, OTP, Ecto, or Phoenix/LiveView code. It corrects the staff-level judgment calls a capable model gets wrong by default — reaching for a GenServer as the universal tool, rescuing instead of letting processes crash, N+1 Ecto access, non-atomic writes, unbounded/linked Task fan-out, atom exhaustion from user input, eager Enum where Stream fits, and calling Repo from the web layer. Applies whenever the work touches process design, error handling, concurrency, data access, or LiveView, even if the user doesn't name them.

# Staff-Level Elixir Distilled staff-level judgment for Elixir on the BEAM — the design decisions Elixir, OTP, Ecto, and Phoenix force, and how an experienced engineer settles them, written so an agent applies them while writing or reviewing code. Each rule corrects a specific wrong default; there is no rule for things the model already gets right (syntax, basic idioms, the standard library). ## When to Apply - Writing or reviewing OTP code — deciding whether something needs a process, and which (GenServer, Task, Agent, ETS, supervisor) - Handling failure — choosing between tagged tuples, exceptions, `with`, and letting a process crash - Writing Ecto — queries, associations, transactions, changesets, large result sets - Building Phoenix controllers and LiveViews — context boundaries, socket state, mount lifecycle - Running concurrent work — parallel fan-out, fire-and-forget tasks, handling untrusted input - Any request to make Elixir code more idiomatic, more fault-tolerant, or "staff-level" ## Rule Categories | # | Category | Prefix | Covers | |---|----------|--------|--------| | 1 | Process & OTP Design | `otp-` | Whether a problem needs a process, and which one; supervision | | 2 | Error Handling & Let-It-Crash | `err-` | Signalling failure; when to raise, rescue, or crash | | 3 | Idioms & Design Choices | `data-` | Pattern matching, Stream/Enum, iolists, pipes, macros | | 4 | Concurrency & the Scheduler | `conc-` | Bounded/isolated parallelism; atom-table safety | | 5 | Ecto & Data Access | `ecto-` | N+1, atomic writes, constraint races, large sets | | 6 | Phoenix & LiveView | `phx-` | Context boundaries, socket memory, mount lifecycle | ## Quick Reference ### 1. Process & OTP Design - [`otp-genserver-not-default`](references/otp-genserver-not-default.md) — a GenServer is a serialization point, not the default abstraction - [`otp-ets-for-shared-reads`](references/otp-ets-for-shared-reads.md) — read-heavy shared state belongs in ETS, not behind a GenServer.get - [`otp-supervise-over-manual-restart`](references/otp-supervise-over-manual-restart.md) — child specs + DynamicSupervisor/Registry, not hand-rolled restart logic - [`otp-handle-continue-init`](references/otp-handle-continue-init.md) — defer heavy startup to `handle_continue`; `init/1` blocks the supervisor ### 2. Error Handling & Let-It-Crash - [`err-tagged-tuples-vs-raise`](references/err-tagged-tuples-vs-raise.md) — `{:ok/:error}` for expected failures, raise only for invariant violations - [`err-let-it-crash`](references/err-let-it-crash.md) — don't rescue to mask bugs; let the supervisor restart clean state - [`err-with-happy-path`](references/err-with-happy-path.md) — chain fallible steps with `with`; keep the error term intact ### 3. Idioms & Design Choices - [`data-pattern-match-over-conditionals`](references/data-pattern-match-over-conditionals.md) — function clauses + guards over if/cond on argument shape - [`data-stream-vs-enum`](references/data-stream-vs-enum.md) — Stream for large/lazy/early-exit, Enum for small concrete lists - [`data-iolists-over-concat`](references/data-iolists-over-concat.md) — build large output as iolists; `<>` in a loop is O(n²) - [`data-pipe-idioms`](references/data-pipe-idioms.md) — pipe a data subject through transformations, not to save a variable - [`data-functions-over-macros`](references/data-functions-over-macros.md) — solve it with a function before reaching for a macro ### 4. Concurrency & the Scheduler - [`conc-async-stream-bounded`](references/conc-async-stream-bounded.md) — `Task.async_stream` with `max_concurrency` + explicit `timeout` - [`conc-task-supervised-nolink`](references/conc-task-supervised-nolink.md) — supervised, unlinked Task for fire-and-forget work - [`conc-atom-exhaustion`](references/conc-atom-exhaustion.md) — never build atoms from external input; use `to_existing_atom` ### 5. Ecto & Data Access - [`ecto-preload-n-plus-one`](references/ecto-preload-n-plus-one.md) — preload up front, never inside an Enum loop - [`ecto-multi-for-transactions`](references/ecto-multi-for-transactions.md) — `Ecto.Multi` for atomic multi-step writes with failure attribution - [`ecto-db-constraints-over-validation`](references/ecto-db-constraints-over-validation.md) — DB constraint + `unique_constraint`, not a racy validation query - [`ecto-stream-large-sets`](references/ecto-stream-large-sets.md) — `Repo.stream`/batch large sets; `Repo.all` OOMs at scale - [`ecto-atomic-counters`](references/ecto-atomic-counters.md) — `update_all` with `inc:` for counters; read-modify-write loses updates ### 6. Phoenix & LiveView - [`phx-context-boundary`](references/phx-context-boundary.md) — web layer calls contexts, never Repo or schemas directly - [`phx-liveview-streams`](references/phx-liveview-streams.md) — streams for large/growing collections, not full lists in assigns - [`phx-mount-twice-connected`](references/phx-mount-twice-connected.md) — gate side effects on `connected?/1`; mount runs twice ## How to Use Read a reference file when its decision comes up. Each rule names the wrong default it corrects, then shows the canonical way (with an incorrect/correct contrast only where the wrong way is a real trap). - [Section definitions](references/_sections.md) — category structure and ordering - [Rule template](assets/templates/_template.md) — for adding new rules - [AGENTS.md](AGENTS.md) — auto-built table of contents across all rules ## Reference Files | File | Description | |------|-------------| | [references/_sections.md](references/_sections.md) | Category definitions and ordering | | [assets/templates/_template.md](assets/templates/_template.md) | Template for new rules | | [metadata.json](metadata.json) | Version and source references |

在 GitHub 查看

返回搜索返回安装

ray-llm

LLM workloads on open-source Ray (pinned to 2.57) — OpenAI-compatible serving with ray.serve.llm (vLLM-backed LLMConfig + build_openai_app) and batch inference with ray.data.llm (build_processor). Corrects the stale defaults a model produces — the archived ray-llm repo and its YAML configs, hand-rolled vLLM engines inside plain Serve deployments, the removed build_llm_processor name, deprecated boolean stage flags, top-level LLMServer/LLMRouter imports, free-form accelerator strings, one-deployment-per-LoRA-adapter designs — with the 2.57 idioms (stage configs, placement_group_config over hand-rolled PGs, deployment_config autoscaling, dynamic LoRA multiplexing, prefix-cache-affinity routing, the full OpenAI endpoint surface). Use when writing, reviewing, or productionizing LLM serving or batch inference on Ray. Classic-ML Ray (Train/Tune/Data/Serve/clusters) lives in the sibling ray skill.

2026-08-15

ray

Production Ray (open-source, pinned to 2.57) for classic-ML workloads from training to serving — Ray Train, Tune, Data, Serve, Core, and cluster deployment on KubeRay. Corrects the older-corpus defaults a model reaches for (ray.air session reporting, Trainer-inside-Tuner, tune.run, map_batches concurrency=, DatasetPipeline/to_torch, max_concurrent_queries, RayServeHandle + ray.get, Deployment.deploy, ray.state, ray.get-in-a-loop) with the 2.57 idioms that replaced them (Train V2 defaults, driver-function tuning, compute strategies, streaming datasets, DeploymentHandle/DeploymentResponse, serve build/deploy, ray.util.state, KubeRay CRDs and Jobs API). Use when writing, reviewing, or productionizing Python code that touches Ray distributed training, data pipelines, hyperparameter tuning, model serving, or Ray cluster operations. LLM serving/batch-inference on Ray lives in the sibling ray-llm skill.

2026-08-15

mlflow-3

MLflow 3 (open-source, pinned to 3.15) for classic-ML MLOps — logging and registering models, promoting versions across dev/staging/prod, standing up a tracking server, evaluating with gates, and serving. Corrects the MLflow 2-era defaults a model reaches for (artifact_path, registry stages and get_latest_versions, top-level mlflow.evaluate with baseline_model, runs-URI registration, pickle serialization, mlruns file stores, MLServer serving) with the MLflow 3 idioms that replaced them (named LoggedModels, aliases and copy_model_version, models.evaluate plus validate_evaluation_results, skops/torch.export defaults, database backends, the FastAPI scoring server). Use when writing, reviewing, or migrating Python code that touches MLflow tracking, the model registry, evaluation, or serving.

2026-08-15

查看这个仓库的全部 Skills

staff-level-elixir

来源信息

安装方式

检查来源文件

同仓库更多 Skills

同仓库更多 Skills