VividKit Crystal clear AI coding
VI
Choose Your Setup Get Started with AgentKit Continue with ClaudeKit Migrate from ClaudeKit to AgentKit Clean cutover from ClaudeKit What Is AgentKit? How AgentKit Works AgentKit Kits, Skills, Agents & Hooks AgentKit Runtime Support AgentKit Desktop App AgentKit Configuration AgentKit CLI Quick Reference AgentKit Skills Cheatsheet AgentKit Workflows AgentKit Statusline Update AgentKit CLI & Kits AgentKit Helper Clean reinstall of AgentKit AgentKit Troubleshooting What is ClaudeKit? ClaudeKit CLI Guide Command Finder Commands Cheat Sheet Workflow Recipes CLI Quick Reference Claude Mechanics Permission Modes ClaudeKit Hooks Coexistence ClaudeKit x Codex CK → Other coding tools How ClaudeKit Works Inside ClaudeKit Finding Your Unknowns Session Recovery Fix from Logs UI/UX Guide Accounts Switcher IDE Config Remote Control Codex App System Diagramming UI Review Gate Promotions Support VividKit

Quick Ref / Command

Engineer Kit

/ak:code-review

Evidence-backed code review

Production-risk review for pending diffs, pull requests, commits, recent context, and codebase scans.

01

Resolve target

02

Load evidence

03

Spec compliance

04

Scout edge cases

Rule 01

KISS, DRY, and YAGNI always.

Rule 02

Resolved review target before opinions.

Rule 03

Evidence before claims; verification before completion.

Rule 04

Polished AI-assisted code can still hide defects.

Execution Map

Core lanes from input to output, aligned with the command's real execution path.

01

Start

Input, scope, route

  1. 1 Resolve target Parse the first matching input: PR number or URL, 7+ char commit hash, --pending, codebase, codebase parallel, or recent context.
  2. 2 Load evidence Fetch the complete PR diff with gh, commit diff with git show, pending staged plus unstaged diff, or the chosen codebase scope.
02

Work

Agent / skill execution

  1. 3 Spec compliance When a plan or specification exists, mark every requirement Pass, Missing, or Extra before quality review.
  2. 4 Scout edge cases Use ak:scout to trace affected files, consumers, data flows, error paths, boundary conditions, and regression risks.
03

Verify

Gate, review, validation

  1. 5 Quality review Run code-reviewer after spec compliance to check standards, security, performance, maintainability, and verification gaps.
  2. 6 Checklist overlay For pre-landing, ship, significant, or security-sensitive changes, apply the base checklist and detected web/API overlays.
04

Close

Report, handoff, artifact

  1. 7 Fix accepted findings Fix Critical immediately and Important before proceeding; re-review cycles stop after three rounds and escalate.
  2. 8 Final verify Run the relevant test, build, lint, or manual reproduction, read the output, and only then make completion claims.

Syntax, arguments, subcommands

How to invoke the skill: syntax, positional arguments, shared options, then each subcommand with its own syntax and outcome.

Syntax

/ak:code-review [#PR | COMMIT | --pending | codebase [parallel]] [--ultra] [--advice] [--yagni]

Arguments

[#PR | COMMIT | --pending | codebase [parallel]]

Review target

Select a pull request, commit, pending workspace, codebase scan, or parallel codebase audit. Omit when recent context already identifies the change or when you want the Skill to ask.

/ak:code-review #123

Shared options

--ultra

Five-reviewer pass

Runs Stage 2 quality review with five independent read-only reviewers and one verifier. Conflicts with codebase parallel.

/ak:code-review --pending --ultra
--advice

Advisory supervision

Adds kongming supervision after spec and quality review, when stuck, and before high-stakes verdicts. It cannot bypass evidence rules.

/ak:code-review #184 --advice
--yagni

Review unneeded scope

Makes findings about cutting unneeded requested scope valid. Without it, requested scope stays a review constraint.

/ak:code-review --pending --yagni

Subcommands

codebase Codebase scan
/ak:code-review codebase [--ultra] [--advice] [--yagni]

Review the full codebase scope with research, edge-case scouting, quality review, verification, and an improvement plan when useful.

--ultra
Use five independent reviewers for the quality-review stage. Do not combine with parallel.
--advice
Ask kongming for counsel at review checkpoints without replacing evidence checks.
--yagni
Include evidence-backed scope-cut findings in the review.

Outcome: A codebase-risk review with confirmed findings, verification gaps, and bounded repair direction.

/ak:code-review codebase --advice
codebase parallel Parallel codebase audit
/ak:code-review codebase parallel [--advice] [--yagni]

Groups edge cases into independent reviewer scopes, then aggregates and verifies blocking findings. It cannot run with --ultra.

--advice
Ask kongming for counsel at review checkpoints without replacing evidence checks.
--yagni
Include evidence-backed scope-cut findings in the review.

Outcome: Independent area reviews joined into one evidence-checked audit with handled and unhandled risks.

/ak:code-review codebase parallel --yagni

Output / publish flags

Only flags that change the artifact or publish it. Not the full option set.

--pending

Pending workspace

Review staged and unstaged local changes against HEAD, including repository status.

Example

/ak:code-review --pending
#PR or URL

Pull request

Fetch PR metadata, changed files, and the full diff with gh before reviewing.

Example

/ak:code-review #123
COMMIT

Commit hash

Validate a 7–40 character hexadecimal commit and review its git show diff.

Example

/ak:code-review abc1234
codebase parallel

Parallel codebase audit

List edge cases, group up to six independent scopes, and verify them with parallel code-reviewer agents.

Example

/ak:code-review codebase parallel

Modes / Routes

#PR or URL

PR review

Research:
gh pr diff full PR
Review:
Spec + quality review
Validate:
Fresh checks before verdict
COMMIT

Commit review

Research:
git show one commit
Review:
Regression and scope review
Validate:
Verify accepted fixes
--pending

Pending workspace

Research:
git diff cached + unstaged
Review:
Local defect scan
Validate:
Fresh local evidence
codebase parallel

Parallel codebase audit

Research:
Scout edge-case categories
Review:
Scoped reviewers per area
Validate:
Aggregate and verify findings

Sample Prompt

Concrete invocations for each flag, subcommand, mode, or route available in the live workflow.

Pending workspace Recommended 
/ak:code-review --pending
Use when:
Review staged and unstaged local work before claiming it is ready.
Expected:
Resolves the pending workspace with status and combined diff, asks for intent if needed, scouts edge cases, reviews quality, and requires fresh verification for accepted fixes.
Pull request 
/ak:code-review #123
Use when:
Review a GitHub pull request by number using authenticated gh access.
Expected:
Fetches PR metadata, changed files, and the full PR diff, then separates spec compliance from production-risk quality findings with evidence.
Commit hash 
/ak:code-review abc1234
Use when:
Review one local commit and its parent context before building on it.
Expected:
Validates the commit, reads its message, changed files, and git show diff, then reviews regressions, scope fit, quality, and verification gaps.
Parallel codebase audit 
/ak:code-review codebase parallel
Use when:
Audit independent codebase areas by first listing edge cases, then assigning reviewer scopes.
Expected:
Identifies edge-case categories, assigns up to six independent code-reviewer scopes, aggregates handled versus unhandled cases, and verifies blocking findings.

Handled Scope

  • Resolve target
  • Load evidence
  • Spec compliance
  • Scout edge cases
  • Quality review
  • Checklist overlay

Related skills

Next