Code Review & Refactoring

Review local diffs or GitHub pull requests for bugs, security gaps and missing tests

Review a stack of pull requests in parallel and post one verdict per change

Run a pre-merge checklist built from patterns in recurring reviewer feedback

Track technical debt from chats and PR reviews and surface it in weekly retros

Review diffs for bugs, injection flaws, N+1 queries and architecture problems

Plan incremental migrations of legacy systems with service boundaries and roadmaps

Make the smallest code change that fixes one well-scoped issue

Find tech debt, TODOs and debug code in recent merges and rank cleanup fixes

Replace lodash imports with es-toolkit to shrink bundles, with compat guidance

Recommend the right es-toolkit utility for a task instead of hand-rolled code

Review code changes or PR diffs for security, performance and correctness issues

Find, categorize and rank technical debt to decide what to refactor first

Verify a diff with an independent reviewer from another vendor and loop until clean

Check a final diff's consumers, focused tests, evidence and gaps before merge

Run automated PR reviewers repeatedly until every finding is settled

Review an existing fix PR against its audited repro and the full diff

Review source changed in a project phase for bugs, security and quality problems

Upgrade package.json dependencies to their latest versions with ncu

Record and analyze Android traces, CPU stacks and heap data to find bottlenecks

Open a browser code review of a worktree or pull request and act on feedback

Navigate a live Hunk diff session and leave inline review comments from the CLI

Audit a codebase read-only and write prioritized plans for other models to run

Scan a working directory for bugs, dead code and security hotspots without edits

Trace requests, dependencies and change impact through a software graph before grep

Run fast Git hooks that lint, format and validate code as a pre-commit drop-in

Review database patches the way PostgreSQL committer Tom Lane would

Phrase code review feedback in Chinese with graded labels and common response templates

Check code review feedback for technical accuracy before acting on it

Request a structured code review of finished work before merging

Instrument Ascend MoE operators with trace points that produce Chrome profiling traces

Convert DeepEP operator calls to CAM operators with HCCL and NPU device changes

Review assigned work at each review boundary in a multi-agent team

Migrate Webiny v5 plugins to the v6 architecture with DI services

Review code for bugs, security and performance with actionable feedback

Run an autonomous review and fix loop over code with a findings ledger

Set up Python projects with uv, ruff and ty instead of pip, Poetry, mypy or black

Loop review and fixes on a branch's changes until its PR review comes back clean

Generate an interactive HTML walkthrough for reviewing code changes

Get an independent code review of changes from Codex or Antigravity

Check code against its spec to find unmet, contradicted and undocumented behavior

Run TypeScript compile checks and show only errors from files you just changed

Run tsc --noEmit with fallbacks and separate new errors from old ones

Audit TypeScript imports for type-only misuse, duplicates and incomplete barrel exports

Run TypeScript type checks with a shell helper when the normal command times out

Write special-character files via heredoc and type-check only the new TypeScript files

Find performance anti-patterns and slow code with parallel specialist audits

Plan and verify dependency upgrades in risk-tiered batches with clean commits

Run ESLint, Biome, Prettier or native formatters with auto-fix and report leftovers

Review staged, unstaged or committed changes for security, bugs and quality