Use Claude to remember family recipe ratings
Chandrika Srivastava wanted weekly meals chosen from recipes her family already knew, with preferences remembered between planning sessions. She downloaded three collections from Google Docs as Markdown files and placed them in a local folder for Claude Code.
Original prompt
Shared by the source author. See source notes for context.
Make a weekly meal plan by choosing 5-6 recipes from the three recipe lists I have.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Chandrika Srivastava
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Local weekly plans plus persistent selection rules and separate child/adult recipe ratings collected through the /rating skill.
- Human involvement
- Srivastava reviewed plans, refined food rules and supplied the family's recipe ratings.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Export existing recipe collections into local Markdown files.
- Refine weekly choices and save preferences in CLAUDE.md.
- Record separate adult and child ratings against each recipe.
Chandrika Srivastava wanted weekly meals chosen from recipes her family already knew, with preferences remembered between planning sessions. She downloaded three collections from Google Docs as Markdown files and placed them in a local folder for Claude Code.
The initial plan selected five or six recipes, with one nonvegetarian dish and the rest vegetarian. Srivastava refined the choices through conversation, including a rule against multiple lentil dishes or multiple salads in the same week. Claude saved the accepted plan as a local file, which she named weekly-meal-plan.md, and turned the accumulated rules into CLAUDE.md.
A second problem was remembering the children's reactions. She asked Claude to create a skill that collected separate adult and child ratings. The resulting /rating interaction asked for recipe names and the two scores, then found each recipe's row in the weekly plan and added the ratings. AskUserQuestion made the process interactive so entries were not silently skipped.
Srivastava continued accepting proposed plans and supplying household feedback. The result was a small system of recipes, persistent selection rules and recorded reactions. She describes it as useful already, without reporting a quantified nutrition improvement, lower grocery bill or an unattended cooking routine.
Prompt
Original prompt
Source: chandrikasrivastava.substack.com · Chandrika Srivastava
Frequently asked questions
How did existing recipes reach the agent?
Srivastava downloaded three Google Docs collections as Markdown files into the project folder.
What preference needed explicit correction?
She told Claude not to pick two lentil dishes or two salads within one week.
Why keep two rating columns?
The skill recorded adult and child reactions separately, preserving the children's preferences rather than relying on memory.
More details
About this account
Claude chose recipes, saved rules and created an interactive rating skill that updated the meal-plan file.
Firsthand source opened and read. Prompt is a verbatim excerpt of the published original instruction.
Credit: Chandrika Srivastava
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Markdown · CLAUDE.md · AskUserQuestion
Family recipes · Child and adult ratings · Meal preferences
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.