Hi Product Hunt ๐
Iโm Jeremy, and today Iโm launching ChefSkills, an open-source framework for giving AI agents more structured culinary reasoning.
The idea came from a simple problem: cooking is not just retrieving a recipe.
Ingredients change. Heat changes them again. Equipment behaves differently. Timing matters. Substitutions create side effects. And when something goes wrong halfway through a cook, the useful answer is rarely just another set of instructions.
ChefSkills is designed to help AI reason about cooking as a changing physical system, identify the current state of a dish, understand why it differs from the intended result, and choose a useful next step.
The public preview currently includes:
๐จโ๐ณ Focused culinary Agent Skills covering core cooking, ingredients, techniques, substitutions, recipe development, scaling, food safety, baking, proteins, sauces, fermentation, equipment and more
๐ง Culinary state modeling for ingredients, transformations, observations, workflow and recovery
๐ Task routing that loads the smallest useful combination of specialist skills instead of treating every culinary question the same
๐งช Evaluation and regression tooling, including 38 evaluated fixtures and hard safety gates
๐ GitHub Copilot installation support through portable skill packages
One thing that matters to me with this project is separating useful AI assistance from pretending AI has professional authority.
ChefSkills is not a culinary credential, food-safety certification, or replacement for current regulatory or professional guidance. It is a framework for making culinary AI behavior more structured, inspectable and testable.
The project is currently public alpha, MIT licensed, and open to contributors.
Iโd especially like feedback from chefs, cooks, culinary educators and AI builders:
What is the hardest part of cooking or kitchen work for an AI assistant to actually understand, rather than simply explain?
GitHub:
https://github.com/jeremylongwor...