本文へ移動
cccskills
無料GitHub で公開

remix-v2-data-flow-review

Reviews Remix v2 loaders and actions for mutations-in-loader, missing validation, leaked server fields, wrong return helpers, v1 useTransition holdovers, and revalidation traps. Use when reviewing loader/action code in a Remix v2 codebase.

インストール方法を見る

含まれるファイル(5)

  • SKILL.md10.0 KB
  • references/actions.md6.5 KB
  • references/defer-await.md4.9 KB
  • references/loaders.md6.2 KB
  • references/revalidation.md6.5 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Remix v2 Data Flow Code Review

Targets TypeScript route modules importing from @remix-run/*. See remix-v2-data-flow for canonical patterns.

Scope

  • In scope: route modules under app/routes/ exporting loader, action, shouldRevalidate, or headers; components that consume useLoaderData, useActionData, useNavigation, useFetcher, useRevalidator, <Await>.
  • Out of scope: form ergonomics (<Form> markup, accessibility, useFetcher UI patterns) → covered by remix-v2-forms-review. Route module conventions, file naming, nested routing, error boundary placement → covered by remix-v2-routing-review.
  • Imports expected: @remix-run/node (or @remix-run/cloudflare / @remix-run/deno) for server utilities; @remix-run/react for hooks and components.

Quick Reference

Issue TypeReference
Mutations in loader, missing validation, leaked server fields, throwing primitives, missing param checksreferences/loaders.md
Unvalidated FormData, json instead of redirect on success, missing error case, leaked actionDatareferences/actions.md
useTransition v1 holdover, missing pending state, blanket shouldRevalidate: false, misused useRevalidatorreferences/revalidation.md
defer for already-fast data, missing <Suspense>, no errorElement on <Await>, awaiting what should streamreferences/defer-await.md

Review Checklist

  • Data needed for first render is in loader, not useEffect
  • Loaders only read; writes live in action
  • request.formData() results are validated (zod/valibot/invariant) before use
  • Loader/action return values are projected DTOs — no password hashes, tokens, or internal_* fields
  • useLoaderData<typeof loader>() uses the type annotation form (not as Foo)
  • 404 / auth short-circuits throw a Response (or json/redirect), never a plain Error or string
  • Successful action returns redirect(...) (PRG); validation failures return json({ errors }, { status: 400 })
  • Action handles both success and error branches; no silent return null
  • params.foo is checked with invariant / zod before use
  • Pending UI reads useNavigation() / fetcher.state — no useTransition
  • formMethod comparisons use UPPERCASE ("POST", not "post")
  • shouldRevalidate returns defaultShouldRevalidate by default; opt-outs are narrow and justified
  • defer() is used only when at least one promise streams (no await before passing it)
  • Every <Await> is wrapped in <Suspense> and has an errorElement
  • useRevalidator().revalidate() is reserved for focus/polling/SSE — not called immediately after a <Form> post or fetcher.submit (Remix already revalidates).

Valid Patterns (Do NOT Flag)

These are correct Remix v2 usage and must not be reported as issues:

  • useEffect for client-only data — Loaders run server-side; localStorage, window dimensions, IntersectionObserver, and browser-only APIs belong in useEffect.
  • loader returning null — A loader may legitimately return null (e.g. optional resource not present); flag only if it should be a 404 throw.
  • useLoaderData<typeof loader>() as type annotation — The <typeof loader> is a generic parameter feeding SerializeFrom<T>, not a as-style type assertion. Do not flag it as "unsafe cast."
  • Bare new Response(body, init) returns — v2 routes may return any Response; json() is an ergonomic wrapper, not a requirement. Non-JSON bodies (binary, text, streams) correctly skip json().
  • return redirect(...) from an action — Both return redirect(...) and throw redirect(...) are legal in actions; throwing is required only from non-action helpers when you want to exit the calling function.
  • loader declared without the request arg — Loaders may destructure only what they need ({ params }, { context }, or () with no args); the unused arg is not a bug.
  • Parent loader revalidated after an unrelated action — This is default Remix behavior, not a smell. Flag only if shouldRevalidate exists and is wrong.
  • Action returning json({ errors }, { status: 400 }) — This is the canonical validation-error pattern (keeps the form route rendered with field errors). Not the same as the "no redirect on success" anti-pattern.
  • useRevalidator for focus / polling / cross-tab sync — These are the documented use cases; only flag manual revalidate() calls that immediately follow a <Form> post or fetcher.submit Remix would already revalidate.
  • SerializeFrom-induced type changes — Date typed as string, Map typed as {} after deserialization is correct wire-format behavior, not a typing bug.

Context-Sensitive Rules

Only flag these issues when the specific context applies:

IssueFlag ONLY IF
Missing loader (using useEffect instead)Data is available server-side and is NOT a browser-only API read
loader returns a raw ORM objectThe object contains fields a reviewer would not paste into a screenshot (passwords, tokens, internal flags)
Action returns json on successThe action is invoked via <Form> causing a URL change — NOT via useFetcher
Missing pending UINo nav.state / fetcher.state reference exists elsewhere in the file driving the same surface
shouldRevalidate returns falseThe body has no condition or never references formAction / currentParams / nextParams
Manual useRevalidator().revalidate()The call follows a Remix-managed mutation (<Form> post, fetcher.submit) — not focus / polling / websocket
defer() usedEvery promise in the defer({...}) payload was already awaited before the call

Hard gates (before writing findings)

Run these in order. Do not draft user-facing findings until every gate passes for the batch you are about to report.

  1. Location evidence — Pass: Each issue lists the repo path to the route module and either a line range or a short verbatim quote from the file you read (not from memory or diff-only guesswork). Loader/action issues without a path to the export async function loader|action are not reportable.

  2. Exemption check — Pass: For each issue, you can state in one line why it is not covered by Valid Patterns (Do NOT Flag). In particular: confirm useEffect is not loading client-only data; confirm a bare Response return is not intentionally non-JSON; confirm a loader returning null is not a legitimate optional read.

  3. Type-annotation vs type-assertion check — Pass: Before flagging an "unsafe cast" on loader/action consumption, confirm the code uses as (assertion) — not useLoaderData<typeof loader>() (annotation) and not useActionData<typeof action>() (annotation). The generic form is the documented safe path and must not be flagged.

  4. v1 holdover check — Pass: Before flagging "missing pending state," grep the file for useTransition, transition.submission, fetcher.type, formMethod === "post" or formMethod==='post' (lowercase, any whitespace/quote variation), and LoaderArgs / ActionArgs. If present, the finding is a v1-holdover migration issue, not a missing-feature issue — label it accordingly.

  5. Protocol — Pass: You completed the Pre-Report Verification Checklist in review-verification-protocol for this review.

When to Load References

Review Questions

  1. Is data needed for first render fetched in a loader, or is it stuck in a useEffect that defeats SSR and revalidation?
  2. Does every loader return a projected DTO, or do raw ORM records (with password, token, internal_* fields) leak to the browser?
  3. Does every action validate request.formData() with a schema before touching the database?
  4. Does the success branch of each action redirect(...) so refresh / back behaves correctly (PRG)?
  5. Is the consumer code using useLoaderData<typeof loader>() (annotation) — not useLoaderData() as Foo (assertion)?
  6. Do any v1 holdovers remain (useTransition, transition.submission, fetcher.type, lowercase formMethod, LoaderArgs / ActionArgs)?
  7. Does shouldRevalidate return a literal false, or does it reach for defaultShouldRevalidate and opt out narrowly?
  8. Is defer() used only when at least one promise is passed unresolved, and is every <Await> wrapped in <Suspense> with an errorElement?

Additional Documentation

Before Submitting Findings

Complete Hard gates (especially gate 5), then report only issues that still pass the review-verification-protocol pre-report checks.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Use when you need to mine a conversation, session transcript, or design discussion for architectural decisions before writing ADRs. Identifies problem-solution pairs, trade-off debates, technology choices, and explicit "[ADR]" tags. Triggers on "what decisions did we make", "extract decisions from this chat", "find the choices in our discussion", or "summarize architectural decisions". Also useful after long planning sessions to capture decisions that were made implicitly. Does NOT write ADR documents — use adr-writing or write-adr for that.

日本語の概要は準備中です。原文の説明を表示しています。

existential-birds/beagle822026年9月16日 更新

Use when writing or formatting an ADR document using the MADR template, applying Definition of Done (E.C.A.D.R.) criteria, or verifying ADR completeness. Triggers on "write the ADR", "format as MADR", "check ADR quality", "mark gaps in ADR". Also triggers when a decision has been extracted and needs to become a document. Does NOT extract decisions from conversations (use adr-decision-extraction) or orchestrate the full extract-confirm-write workflow (use write-adr).

日本語の概要は準備中です。原文の説明を表示しています。

existential-birds/beagle822026年9月16日 更新

Use when auditing an agent codebase against the 12-Factor Agents methodology, reviewing LLM-powered system architecture, or assessing agentic app compliance. Triggers on "analyze agent architecture", "12-factor audit", "how compliant is this agent", or "evaluate this LLM app". Also applies when comparing frameworks or planning agent improvements. Not for quick checklists — this performs deep per-factor codebase analysis with file-level evidence.

日本語の概要は準備中です。原文の説明を表示しています。

existential-birds/beagle822026年9月16日 更新

Vercel AI Elements for workflow UI components. Use when building chat interfaces, displaying tool execution, showing reasoning/thinking, or creating job queues. Triggers on ai-elements, Queue, Confirmation, Tool, Reasoning, Shimmer, Loader, Message, Conversation, PromptInput.

日本語の概要は準備中です。原文の説明を表示しています。

existential-birds/beagle822026年9月16日 更新

Reviews App Intents code for intent structure, entities, shortcuts, and parameters. Use when reviewing code with import AppIntents, @AppIntent, AppEntity, AppShortcutsProvider, or @Parameter.

日本語の概要は準備中です。原文の説明を表示しています。

existential-birds/beagle822026年9月16日 更新

Use when the user wants a cited, structured read of local documents and project knowledge. Triggers on: "analyze these docs", "scan my project for context", "read the docs folder", "summarize what's in .beagle/concepts/", "extract context from docs/", "what's in this folder", "go read everything in X and tell me what's there". Also invoked programmatically by other beagle skills (prfaq-beagle Ignition, brainstorm-beagle reference points, strategy-interview context grounding) via the companion contract. Does NOT trigger on codebase lookups ("find this function", "search the repo"), web research (use web-research), LLM-as-judge evaluation (use llm-judge), or document editing (use humanize-beagle). Produces a written scan plan, parallel-subagent findings, and a cited synthesis report on disk — never inline prose, never unsourced claims.

日本語の概要は準備中です。原文の説明を表示しています。

existential-birds/beagle822026年9月16日 更新

existential-birds のスキルをすべて見る

このスキルの問題を報告する