Skip to content
Neotoxic-offPublic

About

πŸ§‘β€πŸ³ Your AI-powered coding chef crafting clean, secure, and maintainable code with precision

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

Β 

History

60 Commits

Folders and files

Repository files navigation

Chef logo

Chef

Your AI-powered coding chef crafting clean, secure, and maintainable code with precision

Stars Claude Code plugin 34 languages 127 references

Same 13 tasks, same model, a blind reviewer asks for fixes until it merges. Plain Claude: first answer accepted 59%, 46 review rounds per 100 tasks, $0.120 per task with fixes. Chef 1.3: 90%, 13 review rounds per 100 tasks, $0.181. Chef 1.4: 92%, 10 review rounds per 100 tasks, $0.185.

Without chef Chef 1.2 Chef 1.3 Chef 1.4
The reviewer accepts the first answer 59% 87% 90% 92%
Tasks sent back for fixes 41% 13% 10% 8%
Review rounds, as a share of tasks 46% 15% 13% 10% (βˆ’78%)
Lines the reviewer reads, median (lockfiles excluded) 98 231 233 226 (+131%)
Agent time per task, first answer 21 s 42 s 44 s 46 s (+117%)
Agent time per task, fixes and new reviews 29 s 11 s 7 s 7 s (βˆ’76%)
Cost per task, review fixes included $0.120 $0.194 $0.181 $0.185 (+54%)

13 tasks in 11 stacks, 3 runs each, Sonnet 5.5; percentages compare chef 1.4 with plain Claude. Chef 1.2, 1.3 and 1.4 differ by one or two reviews out of 39: read them as equal. The reviewer is a blind Opus session that never knows chef exists; when it would not merge, its required changes go back to the same setup, up to 2 rounds. The code works either way: every first answer passes its hidden tests. Runs cut short by an API or usage limit are refused, not counted; each result names the harness commit and reviewer that produced it, and every scorer is tested in CI on a known-good and a known-bad answer. How it was measured.

Chef 1.4's first answer costs 134% more than a plain one, because it already does what the reviewer would ask for: tests, strict types and linters, pinned versions, configs checked by their tools. Counting the fixes plain Claude needs afterwards, the gap falls to +54%, for 78% fewer review rounds. In agent time the two end level (73 s and 78 s per task until merged, +7%); the difference is on the reviewer's side: 81% fewer tasks come back for another review.

Across versions, on the Python, Docker and Prometheus tasks every version was run on:

Chef across versions, share of first answers a blind reviewer would merge as is: 0.8.1 50%, 0.9.0 44%, 0.10.0 67%, 1.0.0 78%, 1.1.0 78%, 1.2.0 89%, 1.3.0 100%, 1.4.0 100%; plain Claude 44%.

Chef across versions, median cost per task: 0.8.1 $0.167, 0.9.0 $0.184, 0.10.0 $0.180, 1.0.0 $0.155, 1.1.0 $0.145, 1.2.0 $0.156, 1.3.0 $0.169, 1.4.0 $0.173; plain Claude $0.059.

Agent time per task until merged, mean of 39 runs (13 tasks, 3 runs each). Plain Claude: first answer 21 s, first review 23 s, fixes and new reviews 29 s, 41% sent back. Chef 1.1: 57 s, 24 s, 13 s, 15% sent back. Chef 1.2: 42 s, 26 s, 11 s, 13% sent back. Chef 1.3: 44 s, 30 s, 7 s, 10% sent back. Chef 1.4: 46 s, 25 s, 7 s, 8% sent back. Median time of the first answer by version on the Python, Docker and Prometheus tasks: 0.8.1 45 s, 0.9.0 55 s, 0.10.0 46 s, 1.0.0 45 s, 1.1.0 50 s, 1.2.0 41 s, 1.3.0 41 s, 1.4.0 47 s; plain Claude 15 s.

Per token, chef is less cost-effective: $1 buys 8.2 accepted first answers without it and 5.4 with chef 1.4. What it saves is the reviewer's rounds: 1.10 reviews per task instead of 1.46. Once a review round costs more than $0.18 of the reviewer's time (about 11 seconds at $60/h), chef is the cheaper option; at $10 per round, $11.19 per task against $14.72 (βˆ’24%).

Machine cost per task plus a human reviewer's cost per review round, from $0 to $20. Chef 1.4 costs less once a round costs more than $0.18; at $10 per round, plain Claude $14.72, chef 1.3 $11.48 and chef 1.4 $11.19 per task (24% less than plain Claude).

Reviews per task are the first review plus one after each fix. The model prices a round, not the lines in it: chef 1.4's code takes 2.3 times as many lines to read (median 226 against 98, lockfiles excluded).

Most-used stacks

Nine more tasks in the languages and tools developers use most, run on chef 1.4, 3 runs each, with the current reviewer:

Most-used stacks, 9 tasks. Plain Claude: first answer accepted 44%, 74 review rounds per 100 tasks, $0.150 per task with fixes. Chef 1.4: 67%, 37 review rounds per 100 tasks, $0.260 per task with fixes.

First answer accepted, of 3 Node React Bash SQL PHP PowerShell C C++ Kotlin
Without chef 0 3 3 2 1 0 0 0 3
Chef 1.4 3 1 3 2 3 0 3 1 2

Chef is ahead on 4 stacks, behind on 2 and level on 3; with 3 runs per stack, read the total, not a single stack. Every chef first answer passes its hidden test; plain Claude fails 2 of 27 (PHP, a composer.json that is not valid JSON). How it was measured.

Install

/plugin marketplace add Neotoxic-off/chef
/plugin install chef@chef

That's it: chef applies to every coding prompt from now on.

What chef does

  • Asks before guessing: one question when that settles it, the council when it does not. Remembers your answers in .chef.md.
  • Your word wins: an explicit instruction in your prompt overrides chef's conventions and .chef.md, never security or the project folder.
  • Tests every change, and sets up a formatter, a linter and a strict type checker. Warnings get fixed, never hidden.
  • Types every signature, pins dependencies, images and actions.
  • Knows your stack: 127 references (34 languages, frameworks, infra, CI) with the right tools and configs. The 11 stacks the benchmark covers (Python, TypeScript, Go, Rust, Java, C#, Docker, Helm, Terraform, GitHub Actions, Prometheus) are checked by it; the others say which versions they were written for, and that they are not benchmarked.
  • Brings a council of agents for design decisions, and a reviewer for changes. The benchmark measures neither: its tasks are small and run without them.
  • Talks less: short replies, agent reports capped at 20 lines, and the smallest complete change (lean.md, adapted from ponytail, MIT).

Commands

Command
/chef:level lite | full | max How much process: lite keeps every code rule but skips the council and reviewer.
/chef:status Current level, project settings and detected stacks.
/chef:transform <goal> Upgrade or migrate a codebase in small, tested, committed batches.

Spend less

  • /chef:level lite drops the council and the reviewer agent, the biggest cost on large tasks.
  • Turn chef off in one project, for you only: claude plugin disable chef@chef --scope local.
  • Or skip the plugin: copy export/AGENTS.md, chef's 12 core rules, into your project. It works with any agent (Codex, Cursor, Copilot) and gets part of the effect (in the 1.0 benchmark: 56% merged as is against 89% with chef, at $0.084 per task).

Change a rule

Edit skills/chef/SKILL.md or a stack file in skills/chef/references/, then run bash scripts/check.sh. Changes are listed in the CHANGELOG.

About

πŸ§‘β€πŸ³ Your AI-powered coding chef crafting clean, secure, and maintainable code with precision

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Contributors

Languages