One skill. Craftsman's taste. Zero AI slop.
Omakase is a portable quality standard for AI agent work: twelve rules, a critique rubric, and project memory. Trust the chef: state the goal. Nothing mediocre reaches the table.
You state what you want. Omakase decides how to get it done.
Every agent stays narrow and masterful at one thing.
Mediocre work is never acceptable.
Every output must feel like it came from a top-tier expert.
Generic AI patterns are aggressively rejected, in both code and communication. If it looks, feels, or behaves like typical AI output, it has failed.
Implementation, architecture, refactoring, debugging. Runs the work through the shokunin and hands you the finished result with the reasoning attached.
Independent quality enforcement. Harsh, evidence-based review against the rubric. Nothing significant ships without passing this desk.
Memory, decisions, and synthesis. Recaps what shipped, mines conversations for durable preferences, and keeps the project's taste from evaporating between sessions.
@omakase-engineer add rate limiting with backoff
@omakase-critic review the auth module
@omakase-archivist what did I ship since Monday?Does this feel like it was made by a top-tier expert with years of real experience?
Free of generic patterns, fluff, unnecessary comments, defensive code, synthetic tone?
The simplest possible solution that works. No missed chance to reduce complexity?
Respects the project's context, design principles, existing style, and architecture?
Clean and maintainable, with no over-engineering or unnecessary abstraction?
Writing, design, and copy match the exact intended tone and brand voice?
Improves the codebase, rather than adding bloat and scattered logic?
Would we be proud to ship this exactly as-is, with zero revisions?
Fail any check and the work goes back to the kitchen. It never reaches your table.
Humans approve what should be true.
Agents prove it became true.
Run npx omakaseagent init in this repo. Then read the
AGENTS.md and Omakase skill it installs, and adopt the
twelve rules and the critique gate for all work here from
now on.
When you're done, tell me what standard you're now working
under and how I should hand you tasks.One button and the whole kitchen arrives. The ticket tells your agent what to set up, what to read, and the standard to work under from now on.
Paste it into Claude Code, Cursor, or wherever your agent lives. It adopts the rules, remembers your taste, and holds the critique bar before anything significant is served.
State the goal in plain words. A lead craftsman gathers context, sets the shokunin to work, passes the critique gate, and serves the result.
npx omakaseagent init
npx skills add rblalock/omakaseagent
| Harness | Example |
|---|---|
| Cursor / Claude Code | @omakase-engineer add rate limiting with backoff |
| Claude CLI | claude -p --agent omakase-critic "review the auth module" |
| OpenCode | opencode run --agent omakase-engineer "add rate limiting…" |
| Grok | grok --agent omakase-engineer "…" |
No terminal? Omakase works in chat apps too. You get the standard (the twelve rules and the critique gate) applied to whatever you bring: writing, plans, analysis, code. The full kitchen, with leads and project memory, lives in a coding harness.
Download the skillNeeds code execution enabled (Settings → Capabilities)
Your workspace admin may need to turn skills on