agentsclimarketplace

Omakase off

Skill 2389-research/test-kitchen/skills/omakase-off

This skill should be used as the entry gate for build/create/implement requests. Triggers on "build X", "create Y", "implement Z", "add feature", "try both approaches", "not sure which approach". Offers brainstorm-together or omakase (chef's choice parallel exploration) options. Detects indecision during brainstorming to offer parallel exploration.From its SKILL.md

Install
npx -y skills add 2389-research/test-kitchen --skill omakase-off

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.8 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

Omakase-Off

Chef's choice exploration - when you're not sure WHAT to build, explore different approaches in parallel.

Part of Test Kitchen Development:

  • omakase-off - Chef's choice exploration (different approaches/plans)
  • cookoff - Same recipe, multiple cooks compete (same plan, multiple implementations)

Core principle: Let indecision emerge naturally during brainstorming, then implement multiple approaches in parallel to let real code + tests determine the best solution.

Three Triggers

Trigger 1: BEFORE Brainstorming

When: "I want to build...", "Create a...", "Implement...", "Add a feature..."

Present:

Before we brainstorm the details, would you like to:

1. Brainstorm together - We'll explore requirements and design step by step
2. Omakase (chef's choice) - I'll generate 3-5 best approaches, implement them
   in parallel, and let tests pick the winner

Trigger 2: DURING Brainstorming (Indecision Detection)

Detection signals:

  • 2+ uncertain responses in a row on architectural decisions
  • Phrases: "not sure", "don't know", "either works", "you pick", "no preference"

When detected:

You seem flexible on the approach. Would you like to:
1. I'll pick what seems best and continue brainstorming
2. Explore multiple approaches in parallel (omakase-off)

Trigger 3: Explicitly Requested

  • "try both approaches", "explore both", "omakase"
  • "implement both variants", "let's see which is better"

Workflow Overview

PhaseDescription
0. EntryPresent brainstorm vs omakase choice
1. BrainstormPassive slot detection during design
1.5. DecisionIf slots detected, offer parallel exploration
2. PlanGenerate implementation plan per variant
3. ImplementDispatch ALL agents in SINGLE message
4. EvaluateScenario tests → fresh-eyes → judge survivors
5. CompleteFinish winner, cleanup losers

See references/detailed-workflow.md for full phase details.

Directory Structure

docs/plans/<feature>/
  design.md                  # Shared context from brainstorming
  omakase/
    variant-<slug>/
      plan.md                # Implementation plan for this variant
    result.md                # Final report

.worktrees/
  variant-<slug>/            # Omakase variant worktree

Slot Classification

TypeExamplesWorth exploring?
ArchitecturalStorage engine, framework, auth methodYes
TrivialFile location, naming, config formatNo

Only architectural decisions become slots for parallel exploration.

Variant Limits

Max 5-6 implementations. Don't do full combinatorial explosion:

  1. Identify the primary axis (biggest architectural impact)
  2. Create variants along that axis
  3. Fill secondary slots with natural pairings

Critical Rules

  1. Dispatch ALL variants in SINGLE message - Multiple Task tools, one message
  2. MUST use scenario-testing - Not manual verification
  3. Fresh-eyes on survivors - Required before judge comparison
  4. Always cleanup losers - Remove worktrees and branches
  5. Write result.md - Document what was tried and why winner won

Skills Orchestrated

DependencyUsage
brainstormingModified flow with passive slot detection
writing-plansGenerate implementation plan per variant
git-worktreesCreate isolated worktree per variant
parallel-agentsDispatch all variant subagents in parallel
scenario-testingRun same scenarios against all variants
fresh-eyesQuality review on survivors → input for judge
finish-branchHandle winner (merge/PR), cleanup losers

Example Flow

User: "I need to build a CLI todo app."

Claude: [Triggers omakase-off]
Before we dive in, how would you like to approach this?
1. Brainstorm together
2. Omakase (chef's choice)

User: "1"

Claude: [Brainstorming proceeds, detects indecision on storage]

You seem flexible on storage (JSON vs SQLite). Would you like to:
1. Explore in parallel - I'll implement both variants
2. Best guess - I'll pick JSON (simpler)

User: "1"

[Creates plans for variant-json, variant-sqlite]
[Dispatches parallel agents in SINGLE message]
[Runs scenario tests on both]
[Fresh-eyes review on survivors]
[Presents comparison, user picks winner]
[Cleans up loser, finishes winner branch]

What ships with it: 1 file

7.9 KB alongside SKILL.md

references/

Keep looking

Skills are one crate of 326,970. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.