Review

Full code review gate. Runs 7 agents in parallel on uncommitted changes.

Current Changes

!git changes

Full Diff

!git diff HEAD

Instructions

Include the "Current Changes" and "Full Diff" sections above in each agent prompt instead of telling them to run git diff HEAD.

Launch 7 reviewers in parallel using the Agent tool: subagent_type: "code-reviewer", subagent_type: "architect", general-purpose (naming), subagent_type: "backend-engineer", subagent_type: "frontend-engineer", subagent_type: "context-engineer", and subagent_type: "ux-tester".

Code-Reviewer Agent: Anti-Slop

Dispatch using subagent_type: "code-reviewer" via the Agent tool (not Task tool). Prompt:

code

Review uncommitted changes. Review the diff provided and scan all 12 slop categories. Report using Critical/Important/Minor format. If clean: "No slop found."

Architect Agent

Dispatch using subagent_type: "architect" via the Agent tool (not Task tool). Prompt:

code

Review uncommitted changes. Review the diff provided and apply your full review protocol. Report using Critical/Important/Minor format.

Naming Agent

Dispatch using the Agent tool (not Task tool). Prompt:

code

Apply the /naming skill. Review uncommitted changes. Review the diff provided and check all changed identifiers for naming issues. Report using Critical/Important/Minor format. If clean: "No naming issues found."

Backend-Engineer Agent: Simplicity & Elegance

Dispatch using subagent_type: "backend-engineer" via the Agent tool (not Task tool). Prompt:

code

Review uncommitted changes. Review the diff provided and apply your review mode protocol. Check: reinvented wheels, library leverage, YAGNI, complexity creep, approach quality, API regressions. Report using Critical/Important/Minor format. If clean: "Code is appropriately simple."

Frontend-Engineer Agent: User Flows

Dispatch using subagent_type: "frontend-engineer" via the Agent tool (not Task tool). Prompt:

code

Review uncommitted changes. Review the diff provided and apply your user flow testing protocol. Identify affected user flows, trace each one through the code, and report gaps. Report using Critical/Important/Minor format. If clean: "All user flows verified."

Context-Engineer Agent: Ledger

Dispatch using subagent_type: "context-engineer" via the Agent tool (not Task tool). Prompt:

code

Audit Claude.md files against the current uncommitted changes (diff provided). Report using Critical/Important/Minor format:

**Critical:** (architectural change with no Claude.md, Requirements/Boundaries contradicted by changes)
**Important:** (missing ledger entry for significant change, stale requirement or boundary, hierarchy placement issues)
**Minor:** (template compliance gaps, bloated documentation, pruning opportunities)

If clean: "Ledger is up to date."

DO NOT make any changes. Report findings only.

UX Tester Agent: User Experience

Dispatch using subagent_type: "ux-tester" via the Agent tool (not Task tool). Do NOT include the diff — this agent doesn't read code. Instead, translate the diff into features/flows to test. Prompt:

code

Do a complete UX review of the following features/flows affected by the current changes:
[list features/flows derived from the diff, with URLs/routes if identifiable]

Report using Critical/Important/Minor format. If clean: "UX is clean."

Skip if: no dev server is running, changes are backend-only with no UI impact, or affected routes cannot be determined from the diff.

Aggregation

After all subagents complete:

code

# Code Review

## Critical
[issues]

## Important
[issues]

## Minor
[issues]

If all clear: "No issues found."

Gate

•Critical: Block. Do not proceed.
•Important: Report and ask user.
•Minor: Report and continue.