본문으로 건너뛰기
cccskills
무료GitHub에 공개

review-remix-v2

Comprehensive Remix v2 code review with per-area review skills, run in parallel where the agent supports subagents and sequentially otherwise. Detects Remix v2 in package.json, loads relevant review skills, runs verification protocol.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(1)

  • SKILL.md8.6 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Remix v2 Code Review

Arguments

  • --parallel: Hint to fan out per technology area when the agent supports subagents (see Step 5). When unsupported, the review runs sequentially with identical output.
  • Path: Target directory (default: current working directory)

Step 1: Identify Changed Files

git diff --name-only $(git merge-base HEAD main)..HEAD | grep -E '\.(tsx?|jsx?|mjs|cjs|css)$|^(remix|vite)\.config\.'

If nothing is returned, ask the user for an explicit path list before continuing.

Step 2: Detect Remix v2 + Techs

# Detect Remix v2 in package.json (any official adapter)
grep -E '"@remix-run/(react|node|cloudflare|deno|serve)"' package.json | head -3

# Confirm app router layout
ls app/routes/ 2>/dev/null | head -5

# meta exports (route metadata)
grep -rE 'export const meta|export function meta|export let meta' app/ --include="*.tsx" --include="*.ts" -l | head -3

# Sessions / cookies / auth
grep -rE 'createCookieSessionStorage|getSession\(|commitSession\(|destroySession\(' app/ --include="*.ts" --include="*.tsx" -l | head -3

# Form primitives (forms-review)
grep -rE 'useFetcher\(|<Form|useSubmit\(|useNavigation\(' app/ --include="*.tsx" -l | head -3

# Form anti-patterns: native form / manual fetch (forms-review trigger even without Remix imports)
grep -rE '(^|[^.])<form |fetch\(' app/ --include="*.tsx" -l | head -3

# Error boundaries
grep -rE 'export function ErrorBoundary|export const ErrorBoundary|useRouteError|isRouteErrorResponse' app/ --include="*.tsx" -l | head -3

# Headers / streaming / server-only modules
grep -rE 'export const headers|export function headers|defer\(' app/ --include="*.tsx" --include="*.ts" -l | head -3
find app -type f \( -name '*.server.ts' -o -name '*.server.tsx' -o -name '*.client.ts' -o -name '*.client.tsx' \) | head -5

# Prefetch hygiene (perf-ssr-review)
grep -rE 'prefetch=|<PrefetchPageLinks' app/ --include="*.tsx" -l | head -3

# Loaders / actions (data flow)
grep -rE 'export (async )?function (loader|action)|export const (loader|action)' app/ --include="*.tsx" --include="*.ts" -l | head -3

If the package.json check returns nothing, stop and tell the user: this skill expects a Remix v2 project.

Step 3: Load Verification Protocol

Load the review-verification-protocol skill before any substantive judgment on code. This is the canonical cross-plugin protocol; do not substitute a framework-specific copy.

Pass before Step 5: The skill is loaded (or its checklist is open in context). Do not classify severity or write findings until this gate clears.

Step 4: Load Skills

Read each applicable skill below (open its SKILL.md) so its guidance is in context before you review that area.

Always load (a non-trivial Remix v2 app exercises all six areas):

Detection telemetry (records what each area review will actually find; does not gate loading — all six skills always load, but record matches/non-matches for the final report's coverage section):

ConditionSkill
app/routes/ presentremix-v2-routing-review
loader / action exports foundremix-v2-data-flow-review
<Form>, useFetcher, useSubmit, useNavigation, native <form, or manual fetch( foundremix-v2-forms-review
ErrorBoundary / useRouteError foundremix-v2-error-boundaries-review
headers export, defer(, or .server.ts / .client.ts files or prefetch=/<PrefetchPageLinks foundremix-v2-perf-ssr-review
meta export or session/cookie helpers foundremix-v2-meta-sessions-review

If a detection row returns no matches, record that explicitly in the report's coverage section — the skill still loads, but the reviewer should note the area had no surface to review.

Step 5: Review

If the agent supports subagents (and --parallel is requested or appropriate), dispatch one subagent per area in parallel; otherwise run the same areas sequentially in a single context. Both paths produce identical output.

Parallel path:

  1. Detect all areas upfront
  2. Dispatch one subagent per area, each loading its skill and reviewing its domain only
  3. Wait for all subagents to return
  4. Consolidate findings into a single output

Sequential path:

  1. Load applicable skills
  2. Review routing + nested route structure first
  3. Review loader/action data flow
  4. Review forms and progressive enhancement
  5. Review error boundaries
  6. Review meta, sessions, and headers/SSR
  7. Consolidate findings

Step 6: Verify Findings

Before reporting any issue:

  1. Re-read the actual code (not just diff context)
  2. For "unused" claims — did you search all references, including route-file conventions?
  3. For "missing" claims — did you check parent routes, root.tsx, or framework defaults?
  4. For syntax issues — did you verify against Remix v2 docs (not v1, not React Router v7)?
  5. Remove findings that are style preferences, not actual issues

Pass before promoting to Critical/Major: For that item, (2)–(4) are satisfied with a concrete artifact — opened file at FILE:LINE, grep output for references, or cited parent/framework code — not only diff context.

Step 7: Review Convergence

  • Single-pass: report ALL issues across routing, data flow, forms, errors, SSR/perf, meta/sessions, types, and security in one pass. Before submitting, ask whether your recommended fixes would themselves trigger new findings — if yes, include those now.
  • Scope: review only code in the diff and directly related existing code. Missing test coverage is ONE Minor issue, no implementation prescriptions. No new dependencies.
  • Fix complexity: real severity for fixes to existing code. Requests for net-new code (new modules, dependencies, test suites, extracted abstractions) are Informational, not blocking.
  • Re-review: verify previous fixes only — no new discovery, no re-flagging unfixed Minor/Nice-to-Have items.

Output Format

## Review Summary
[1-2 sentence overview]

## Issues

### Critical (Blocking)
1. [FILE:LINE] ISSUE_TITLE
   - Issue / Why / Fix

### Major (Should Fix)
2. [FILE:LINE] ISSUE_TITLE
   - Issue / Why / Fix

### Minor (Nice to Have)
N. [FILE:LINE] ISSUE_TITLE
   - Issue / Why / Fix

### Informational (For Awareness)
N. [FILE:LINE] SUGGESTION_TITLE
   - Suggestion / Rationale

## Good Patterns
- [FILE:LINE] Pattern description (preserve this)

## Verdict
Ready: Yes | No | With fixes 1-N (Critical/Major only; Minor items are acceptable)
Rationale: [1-2 sentences]

Post-Fix Verification

After fixes are applied, run:

npm run lint
npm run typecheck
npm run test

All checks must pass before approval.

Gates

Advance in order; do not skip a pass condition by restating it informally.

  1. Scope recorded — Pass when: You have the output of the Step 1 command (or an explicit substitute path list) naming what is in scope.
  2. Protocol + always skills loaded — Pass when: review-verification-protocol and all six remix-v2-*-review skills are loaded before the first severity judgment.
  3. Conditional skills audited — Pass when: For each Step 2 detection row you either confirmed a match or recorded that the command returned no results.
  4. Critical/Major evidence — Pass when: Each such finding cites a FILE:LINE that exists in the tree and meets the Step 6 pass rule for that finding type.
  5. Single output — Pass when: The Issues section uses one continuous numbering sequence and the deliverable satisfies Step 7 single-pass completeness.

Rules

  • Load skills BEFORE reviewing — not after.
  • Number every issue sequentially with FILE:LINE and clear Issue / Why / Fix.
  • This is Remix v2 — do not apply React Router v7, Remix v1, or Next.js ("use client", server components) conventions.
  • Run the Post-Fix Verification block after fixes; Verdict ignores Minor and Informational items.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

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.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

existential-birds/beagle842026년 9월 16일 업데이트

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).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

existential-birds/beagle842026년 9월 16일 업데이트

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.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

existential-birds/beagle842026년 9월 16일 업데이트

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.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

existential-birds/beagle842026년 9월 16일 업데이트

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

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

existential-birds/beagle842026년 9월 16일 업데이트

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.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

existential-birds/beagle842026년 9월 16일 업데이트

existential-birds의 스킬 모두 보기

이 스킬의 문제 신고