https://github.com/simonepri/let-em-cook
🤌🏻 Teach AI how to cook (code) the way I like it.
https://github.com/simonepri/let-em-cook
Last synced: 2 months ago
JSON representation
🤌🏻 Teach AI how to cook (code) the way I like it.
- Host: GitHub
- URL: https://github.com/simonepri/let-em-cook
- Owner: simonepri
- License: mit
- Created: 2026-03-14T23:53:45.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-03-15T00:47:18.000Z (4 months ago)
- Last Synced: 2026-03-15T12:13:20.008Z (4 months ago)
- Language: Shell
- Size: 418 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: readme.md
- License: license
- Agents: AGENTS.md
Awesome Lists containing this project
README
🤌🏻 Teach AI how to cook (code) the way I like it.
## Synopsis
Out of the box, AI writes bland code — generic, safe, unseasoned. This repo encodes engineering taste and development workflows so any AI coding assistant follows my style.
## Philosophy
AI is already smart. What it lacks is _taste_ — your preferences, your reasoning style, your workflow. This repo teaches it three things:
1. **Principles** — code style priorities, what good looks like, what to avoid
2. **Reasoning** — how to think about any change: Problem → Goal → Solution
3. **Workflow** — how work moves from idea to shipped code
Every rule and skill is written by a human, for a human — with parenthetical hints for agents where needed. They read like team guidelines, not robot instructions.
## How `/cook` works
`/cook` is the full autopilot. You describe what you want, then walk away.
```
You: /cook add rate limiting to the API
┌──────────┐
│ /plan │ architect, interfaces, phases
└────┬─────┘
▼
┌──────────┐
│ ask user │ confirm plan — or /research gaps, then revise
└────┬─────┘
▼
┌──────────┐
│ /execute │ phase 1 → phase 2 → …
└────┬─────┘
▼
┌──────────┐
│ verify │ smoke-test the result
└────┬─────┘
▼
┌──────────┐
│ /commit │ stage + commit
└────┬─────┘
▼
┌──────────┐
│ /polish │ review + /amend → 3x LGTM
└────┬─────┘◄── design gaps loop back to /plan, else creates report
▼
┌──────────┐
│ ask user │ approve, or back to /plan
└────┬─────┘
▼
┌──────────┐
│ /pr │ push + open PR
└──────────┘
```
Each phase runs in an isolated subagent. The orchestrator stays clean, passes only what the next step needs, and can reset any phase that goes off track.
## Commands
| Command | What it does |
| ----------- | ------------------------------------------------------------------------ |
| `/cook` | Full autopilot — plan, execute, polish end-to-end |
| `/research` | Investigate libraries, APIs, and approaches before planning |
| `/plan` | Architecture-level design — interfaces, data flow, parallelizable phases |
| `/execute` | Execute a plan file — spawn parallel subagents per phase |
| `/commit` | Stage changes + conventional commit message |
| `/amend` | Amend the last commit safely |
| `/review` | Two-phase code review — investigate, then report confirmed issues only |
| `/polish` | Iterative review+fix loop — 3 consecutive LGTMs to converge |
| `/pr` | Create a PR — description aggregated from commits |
| `/diff` | Fetch diffs for any scope (PR, commit, branch, staged, working tree) |
| `/fork` | Snapshot context to a temp file — explore a rabbit hole, resume later |
## Installation
```bash
git clone https://github.com/simonepri/let-em-cook.git /tmp/let-em-cook
/tmp/let-em-cook/install.sh .
rm -rf /tmp/let-em-cook
```
The script copies rules and skills into `.agents/`, sets up `AGENTS.md` with the command table, and creates symlinks for Claude Code (`.claude/` → `.agents/`, `CLAUDE.md` → `AGENTS.md`). It asks before overwriting anything that already exists.
Run the same command again to update to the latest version.
## Structure
After installation, your project will have:
```
.agents/
rules/
code.md # Code style principles
workflow.md # Reasoning framework + focus management
skills/
cook/SKILL.md # Full autopilot
research/SKILL.md # Pre-planning investigation
plan/SKILL.md # Architecture-level design
execute/SKILL.md # Plan execution engine
commit/SKILL.md # Commit workflow
amend/SKILL.md # Amend workflow
review/SKILL.md # Code review
polish/SKILL.md # Review+fix loop
pr/SKILL.md # PR creation
diff/SKILL.md # Diff fetching
fork/SKILL.md # Context snapshot for branching
AGENTS.md # Entry point (works with any AI tool)
CLAUDE.md → AGENTS.md # Claude Code compatibility
.claude/
rules → ../.agents/rules # Claude Code native discovery
skills → ../.agents/skills
```
## Customization
Edit `.agents/rules/code.md` to match your style. Add your own skills to `.agents/skills/`. Delete any commands you don't use.
---
_Let 'em cook._ 🔥