Comprehensive multi-stage code review using specialized subagents
View on GitHubFradSer/dotclaude
review
January 24, 2026
Select agents to install to:
npx add-skill https://github.com/FradSer/dotclaude/blob/main/review/skills/hierarchical/SKILL.md -a claude-code --skill hierarchicalInstallation paths:
.claude/skills/hierarchical/# Hierarchical Code Review ## Context - Current branch: !`git branch --show-current` - Git status: !`git status --porcelain` - Base branch: !`(git show-branch | grep '*' | grep -v "$(git rev-parse --abbrev-ref HEAD)" | head -1 | sed 's/.*\[\([^]]*\)\].*/\1/' | sed 's/\^.*//' 2>/dev/null) || echo "develop"` - Changes since base: !`BASE=$(git merge-base HEAD develop 2>/dev/null || git merge-base HEAD main 2>/dev/null) && git log --oneline $BASE..HEAD` - Files changed since base: !`BASE=$(git merge-base HEAD develop 2>/dev/null || git merge-base HEAD main 2>/dev/null) && git diff --name-only $BASE..HEAD` - Test commands available: !`([ -f package.json ] && echo "npm/pnpm/yarn test") || ([ -f Cargo.toml ] && echo "cargo test") || ([ -f pyproject.toml ] && echo "pytest/uv run pytest") || ([ -f go.mod ] && echo "go test") || echo "no standard test framework detected"` ## Requirements - Use **@tech-lead-reviewer** — architectural impact assessment — to scope architectural risk before launching specialized reviews. - Run parallel reviews with: - **@code-reviewer** — logic correctness, tests, error handling. - **@security-reviewer** — authentication, data protection, validation. - **@ux-reviewer** — usability and accessibility (skip if purely backend/CLI). - Consolidate findings by priority (Critical → High → Medium → Low) and confidence (High → Medium → Low). - Offer optional implementation support and ensure commits follow Git conventions (详见 `skills/references/git-commit-conventions.md`). ## Your Task **IMPORTANT: You MUST use the Task tool to complete ALL tasks.** 1. Perform a leadership assessment with **@tech-lead-reviewer** — architectural impact assessment — to map risk areas and determine which specialized agents to involve. 2. Launch the required specialized reviews in parallel via the Task tool, collect outcomes, and resolve conflicting feedback. 3. Present a consolidated report with prioritized recommendations, ask whether the user wants fixes implem