PR size-based review depth, performance review checklist, architecture conformance checks, and framework-specific review patterns.
npx skills add https://github.com/vibeeval/vibecosystem --skill diff-review-strategy
| Category | Lines Changed | Review Depth | Action |
|----------|--------------|--------------|--------|
| XS | 1–10 | Quick scan | Auto-approve if tests pass, typo/doc fixes |
| S | 11–50 | Focused | Check edge cases, naming, one logic path |
| M | 51–200 | Thorough | Full logic review, design check, test coverage |
| L | 201–500 | Walkthrough | Request author explanation, check design first |
| XL | 500+ | Split required | Block merge, ask to split into logical units |
[ ] Does the change do exactly what the title says?
[ ] Are edge cases handled (null, empty, out-of-range)?
[ ] Are variable names clear?
[ ] Are tests updated or added?
[ ] Is the design the simplest solution?
[ ] Are error paths handled?
[ ] Is there duplication that should be extracted?
[ ] Does it follow existing patterns in the codebase?
[ ] Are there security implications (user input, auth)?
[ ] Is the test coverage meaningful (not just happy path)?
1. Read the PR description and linked ticket first
2. Review architecture/design before line-by-line reading
3. Request a walkthrough if intent is unclear
4. If XL: comment "Please split by [feature / layer / file]"
[ ] N+1 query? (loop calling DB inside a loop)
[ ] Missing index on filtered/sorted column?
[ ] SELECT * where only specific columns needed?
[ ] Missing pagination on list endpoints?
[ ] Transaction missing on multi-step writes?
[ ] Unnecessary re-renders? (missing memo/useCallback/useMemo)
[ ] Large component re-renders on every keystroke?
[ ] Images missing width/height (layout shift)?
[ ] Heavy dependency imported at top level (should be lazy)?
[ ] Bundle size impact? (check with next build or vite --report)
[ ] Synchronous I/O inside async handler?
[ ] Missing cache for repeated identical queries?
[ ] Large payload returned when only summary needed?
[ ] Polling where event/webhook would be better?
Controller → Service → Repository → DB (ALLOWED)
Controller → Repository → DB (VIOLATION: skip service)
Service → Controller (VIOLATION: wrong direction)
Repository → Service (VIOLATION: wrong direction)
Red flags in the diff:
import { db } from '../db' inside a route handler fileimport { Router } from 'express' inside a service filefetch() calls inside a React component (should be in a hook or service)Entities (innermost) — no imports from outer rings
Use Cases — import Entities only
Adapters — import Use Cases and Entities
Frameworks — import everything, imported by nothing
[ ] Does feature A import from feature B's internals?
[ ] If yes, should this be a shared utility or event instead?
[ ] Are feature-specific types leaking into shared modules?
[ ] Hooks called conditionally? (rules of hooks violation)
[ ] Missing key prop in lists? Or key is array index?
[ ] useEffect missing cleanup return for subscriptions/timers?
[ ] State mutation instead of new object? ({ ...prev, field: val })
[ ] Prop drilling 3+ levels deep? (consider context or composition)
[ ] Data fetching happening client-side when it could be server-side?
[ ] 'use client' added unnecessarily to a component with no interactivity?
[ ] Large third-party lib imported in a Server Component?
[ ] Dynamic route missing generateStaticParams for static generation?
[ ] API route missing input validation?
[ ] Middleware added in wrong order? (auth before body-parser?)
[ ] Error handler missing 4-argument signature (err, req, res, next)?
[ ] Async handler missing try/catch or asyncHandler wrapper?
[ ] User input used directly in SQL/file path/shell command?
[ ] Missing rate limiter on public endpoints?
nit: Consider renaming `data` to `userProfile` for clarity — easy to
confuse with the raw API response above.
suggestion: This could use `Promise.all` to run both requests in
parallel instead of sequentially. Would save ~300ms per call.
// instead of:
const a = await fetchA()
const b = await fetchB()
// consider:
const [a, b] = await Promise.all([fetchA(), fetchB()])
blocker: This passes `req.params.id` directly into the SQL query
string — SQL injection risk. Use a parameterized query:
db.query('SELECT * FROM users WHERE id = $1', [req.params.id])
nice: Clean separation of concerns here. The service layer staying
unaware of the HTTP layer makes this easy to test.
Control Cursor AI code editor via CLI. Open files, folders, diffs, and manage extensions.
Comprehensive checklist for conducting thorough code reviews covering functionality, security, performance, and maintainability
Systematically evaluate architecture decisions, document trade-offs, and select appropriate patterns. This skill should be used when the user asks about 'architecture decision', 'ADR', 'design pattern selection', 'technology choice', or needs to evaluate architectural trade-offs. Keywords: architecture, ADR, patterns, trade-offs, technical debt, quality attributes, decision record.
Facilitates deliberate skill development during AI-assisted coding. Offers interactive learning exercises after architectural work (new files, schema changes, refactors). Use when completing features, making design decisions, or when user asks to understand code better. Triggers on "learning exercise", "help me understand", "teach me", "why does this work", or after creating new files/modules. Do NOT use for urgent debugging, quick fixes, or when user says "just ship it".
Organize project files and folders for maintainability and scalability. Use when structuring new projects, refactoring folder structure, or establishing conventions. Handles project structure, naming conventions, and file organization best practices.
Automatically trigger review agents after task completion. Use when strategic-planner finishes planning tasks (calls plan-consultant) or when main agent completes coding tasks in /implement workflow (calls code-reviewer). Triggers on phrases like "plan complete", "implementation done", "coding finished", "ready for review".
Provides a structured workflow for planning and executing code reviews like a senior engineer. Use when asked to review code, PRs, or plan a code review task.
Recall project-first and global-supplement ChatCrystal memories before substantive implementation, refactoring, migration, configuration, investigation, or optimization work. Use when the task is non-trivial, has repository or project context, and prior fixes, decisions, pitfalls, or reusable patterns may change the approach.
Take vibeeval/diff-review-strategy from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.