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

Marketing Kit

/ak:cook

Smart feature implementation

Implements known-scope tasks or plans through brainstorm contract, intent detection, scout, research, plan, review, implementation, testing, finalization, and journal steps.

01

Capture contract

02

Detect intent

03

Scout codebase

04

Research and plan

Rule 01

KISS and DRY: implement the requested scope cleanly without optional expansion.

Rule 02

Default mode is interactive; mode flags change workflow shape, not the need for a contract and plan.

Rule 03

Verification evidence is part of the deliverable, not an afterthought.

Execution Map

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

01

Start

Input, scope, route

  1. 1 Capture contract State outcome, constraints, non-goals, and observable acceptance criteria before planning or coding.
  2. 2 Detect intent Classify default interactive mode or requested fast, parallel, auto, no-test, or existing-plan execution.
02

Work

Agent / skill execution

  1. 3 Scout codebase Find project type, relevant files, conventions, docs, plans, and public contracts before detailed planning.
  2. 4 Research and plan Research as needed, write a plan, and pass review gates unless the selected mode changes approval behavior.
03

Verify

Gate, review, validation

  1. 5 Implement scope Implement the full requested scope by default; only --yagni opts into challenging and cutting unnecessary scope.
  2. 6 Review and test Run mandatory review and testing paths, with --no-test only downgrading test execution to a surfaced risk.
04

Close

Report, handoff, artifact

  1. 7 Verify side effects Confirm acceptance criteria, related tests, business logic touchpoints, lint/type/build health, and public contracts.
  2. 8 Finalize Sync plan state, assess docs impact, update live tracking, offer commit workflow, and write journal unless skipped.

Syntax, arguments, subcommands

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

Syntax

/ak:cook [task|plan-path] [--interactive|--fast|--parallel|--auto|--no-test] [--tdd] [--advice]

Arguments

[task|plan-path]

Task or plan path

A concrete technical task, or an approved plan.md path. A plan path loads that plan and skips creating another.

/ak:cook ./plans/utm-capture/plan.md

Shared options

--interactive

Full workflow

Default. Full workflow with human review gates. Still requires the opening contract, plan, implementation, testing, review, and finalization.

--fast

Skip research

Skips research and moves from scout to a concise plan. A plan, tests, and review are still required.

/ak:cook "Add consent-aware UTM capture to the signup form" --fast
--parallel

Concurrent phases

Allows independent phases to execute concurrently. Requires explicit dependencies and non-overlapping file ownership.

--auto

Fewer review pauses

Removes routine human review pauses. Tests, code review, and security policy still apply. Does not authorize publish, deploy, or Git.

/ak:cook ./plans/utm-capture/plan.md --auto
--no-test

Skip tests

Skips the testing stage. Code review stays mandatory and missing test evidence must be warned.

--tdd

Tests-first phases

Adds tests-first behavior within each relevant phase. Does not replace the selected base mode.

/ak:cook "Refactor auth middleware" --tdd
--advice

Advisory checkpoints

Adds kongming advisory checkpoints. Advice does not implement, approve, or expand authority.

/ak:cook "Add consent-aware UTM capture" --advice

Composable flags

How options work together on one invocation. This does not replace the option cards above.

Composable flags documented by the skill: --tdd for tests-first refactoring, --advice for kongming supervision, --yagni to cut unnecessary scope, and --skip-journal to skip the automatic journal step.

Modes / Routes

--interactive

Full workflow

Research:
Yes
Review:
Human review gates
Validate:
Testing required
--fast

Skip research

Research:
No
Review:
Plan still required
Validate:
Testing required
--parallel

Multi-agent execution

Research:
Optional
Review:
Review gates remain
Validate:
Testing required
--auto

Auto-approve steps

Research:
Yes
Review:
Review-cycle rules
Validate:
Testing required
--no-test

Skip testing step

Research:
Yes
Review:
Risk must be surfaced
Validate:
Tests skipped warning

Don't skip the required step

Each item contrasts a tempting shortcut with the required approach.

Tempting shortcut

This is simple enough to code directly.

Do this instead

The skill says simple tasks are where unexamined assumptions waste the most time.

Tempting shortcut

Passing one test means it is done.

Do this instead

Side effects, public contracts, shared tests, and lint/type/build health are part of done.

Sample Prompt

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

Default implementation Recommended 
/ak:cook Add user authentication to the app
Use when:
Use when scope is known but should go through the default interactive workflow.
Expected:
Brainstorm contract, scout, plan, implementation, review, test, and finalize path tied to the requested feature.
Fast path 
/ak:cook Add user authentication to the app --fast
Use when:
Use when the user wants to skip research but still needs scouting, planning, code, and tests.
Expected:
Scout → plan → code flow with required validation gates before the implementation is considered complete.
Execute plan 
/ak:cook path/to/plan.md --auto
Use when:
Use when an accepted plan path should be executed continuously.
Expected:
Execution starts from the plan file, preserves auto mode review cycle behavior, and records phase status updates.
Tests first 
/ak:cook Refactor auth middleware --tdd
Use when:
Use when refactoring should preserve current behavior through tests-first work.
Expected:
Tests for current behavior before refactor, then verification after implementation confirms the contract still holds.
Interactive gates 
/ak:cook Add search to the catalog --interactive
Use when:
Use when the default interactive review gates should stay on.
Expected:
Runs the interactive cook workflow with human review gates, then implements, tests, and finalizes the requested feature.
Parallel agents 
/ak:cook Split billing and notifications work --parallel
Use when:
Independent implementation slices can run in parallel.
Expected:
Groups non-overlapping work, dispatches parallel implementation, then joins review, tests, and finalize for the requested slices.
Skip tests explicitly 
/ak:cook Update README badges --no-test
Use when:
The change has no testable behavior and tests must be skipped on purpose.
Expected:
Records the explicit --no-test exception, still scouts and plans, and completes docs-only implementation without inventing tests.

Handled Scope

  • Capture contract
  • Detect intent
  • Scout codebase
  • Research and plan
  • Implement scope
  • Review and test

Next