Tandoor Recipe CLI

Tandoor CLI Skill

This skill provides AI agents with the ability to interact with a Tandoor Recipe Manager instance using the tandoor-cli command-line tool.

Prerequisites

The tandoor-cli tool must be installed and configured.

Version Compatibility: This skill documentation corresponds to the version specified in the metadata above. Before using this skill:

1. Verify your installed CLI version: tandoor -V
2. Ensure the installed version matches the skill version to avoid unexpected behavior or missing commands
3. If versions don't match, reinstall the CLI tool following the setup instructions in references/SETUP.md

If tandoor -V produces no valid version, see the setup instructions in references/SETUP.md.

Security & Permission Model

⚠️ IMPORTANT: This skill provides mutation authority over your Tandoor instance.

• Read operations (list, search, get, random) are safe and require no confirmation
• Write operations (add, update, import) modify data but are typically reversible
• Destructive operations (delete, clear, household management) require explicit user approval before execution
• Bulk operations (shopping check --all, clear) affect multiple items and require confirmation
• Administrative operations (household management, user assignment, invite creation) require privileged credentials and explicit approval

Token Security:

• Use the least-privileged token possible for your use case
• Prefer read-only tokens if you only need to query recipes
• Avoid using space-owner or admin tokens unless household management is required
• Consider short-lived tokens (days/weeks) instead of long-lived tokens (years)
• Store tokens securely and rotate them regularly
• Revoke tokens immediately if compromised or no longer needed

Supply Chain:

• This skill uses the tandoor-cli npm package
• Verify the package source before first use: https://www.npmjs.com/package/tandoor-cli
• Review the package repository: https://github.com/dcenatiempo/tandoor-cli
• Recommended: Pin to the specific version matching this skill (see version in metadata above) to avoid unexpected updates
• Install pinned version: npm install -g tandoor-cli@<version> (replace <version> with the skill version)

Invocation

tandoor <command> [options]

Output formats and deprecated flags

Commands that print data accept --format text|json|api (default: text):

Deprecated (output): --json with no file path — alias for --format api; stderr warning. Do not use in new agent workflows.

Recommended (input): tandoor add --file recipe.json, tandoor update <id> --file patch.json

Deprecated (input): tandoor add --json <file>, tandoor update <id> --json <file> — still work with stderr warning.

Commands

Read Operations (Safe)

Recipes:

Meal Plans:

Shopping List:

Food Ingredients:

Grocery Categories:

Cook Logs:

Households & Users:

Write Operations (Require Confirmation)

Destructive Operations (Require Explicit Confirmation)

Administrative Operations (Require Privileged Token + Explicit Confirmation)

Important: Tandoor uses households to organize users and recipes. Users are added to households via invite links — you cannot create users directly via the API.

Permission Requirements: Household management commands require special permissions in Tandoor. Most operations need admin/staff privileges. The household invite create command specifically requires space owner authentication — even superusers or staff members will receive a 403 Permission Denied error if they're not the space owner. If you encounter permission errors, you must use the space owner's API token or contact your Tandoor administrator.

Read commands accept --format json (slim, where supported) or --format api (raw). Deprecated: --json (same as --format api).

Agent Behavior Rules

Before executing any command, the agent MUST:

1. For read operations: Proceed without confirmation
2. For write operations: Describe the action and wait for user approval
3. For destructive operations:
   • Clearly explain what will be deleted/modified
   • Warn that the action cannot be easily reversed
   • Wait for explicit user confirmation (e.g., "yes, delete recipe 42")
4. For bulk operations:
   • State how many items will be affected
   • Wait for explicit confirmation
5. For administrative operations:
   • Verify the user has provided an admin/space-owner token
   • Explain the scope of the change (e.g., "this will move user X to household Y")
   • Wait for explicit confirmation

The agent MUST NOT:

• Execute delete, force, bulk, household, invite, or user-assignment commands without explicit user approval
• Assume the user wants destructive actions even if implied by context
• Use --force flags without explicit user instruction
• Log or display the TANDOOR_API_TOKEN value in any output

Configuration

The CLI supports three configuration methods (in precedence order):

1. Environment variables — TANDOOR_URL and TANDOOR_API_TOKEN in the current shell
2. Config file — ~/.config/tandoor-cli/config.json (created by tandoor configure)
3. .env file — a .env file in the current working directory

Run tandoor configure once to save credentials interactively.

Examples

Configure credentials:

tandoor configure

Read example — list recipes as JSON:

tandoor list --limit 5 --format api

Read example — get recipe for editing:

tandoor get 42 --format json

Write example — create a recipe from a JSON file:

# Agent should first ask: "I will create a new recipe from recipe.json. Proceed?"
# Only after user confirms:
tandoor add --file recipe.json

Recipe JSON Schema

When creating recipes with tandoor add --file <file> (or deprecated --json <file>), the JSON file must conform to the following structure:

TypeScript Definition:

interface RecipeCreatePayload {
  name: string;                    // Required: Recipe name
  description?: string;            // Optional: Recipe description
  servings?: number;               // Optional: Number of servings
  working_time?: number;           // Optional: Active cooking time in minutes
  waiting_time?: number;           // Optional: Passive time (baking, marinating) in minutes
  steps: StepCreatePayload[];      // Required: At least one step
}

interface StepCreatePayload {
  instruction: string;             // Required: Step instructions
  order: number;                   // Required: Step order (1, 2, 3, ...)
  ingredients: IngredientCreatePayload[];  // Required: Can be empty array
}

interface IngredientCreatePayload {
  food: { name: string };          // Required: Ingredient name
  unit: { name: string } | null;   // Optional: Unit of measurement (null if not applicable)
  amount: number;                  // Required: Quantity
  note?: string;                   // Optional: Additional notes (e.g., "finely chopped")
  order?: number | null;           // Optional: Order within the step
}

Example Recipe JSON:

{
  "name": "Scrambled Eggs",
  "description": "Quick and easy scrambled eggs",
  "servings": 2,
  "working_time": 5,
  "steps": [
    {
      "instruction": "Crack eggs into a bowl, add milk and salt. Whisk until well combined.",
      "order": 1,
      "ingredients": [
        {
          "food": { "name": "eggs" },
          "unit": { "name": "whole" },
          "amount": 4
        },
        {
          "food": { "name": "milk" },
          "unit": { "name": "tbsp" },
          "amount": 2
        },
        {
          "food": { "name": "salt" },
          "unit": null,
          "amount": 1,
          "note": "to taste"
        }
      ]
    },
    {
      "instruction": "Heat butter in a pan over medium heat. Pour in egg mixture and stir gently until cooked to desired consistency.",
      "order": 2,
      "ingredients": [
        {
          "food": { "name": "butter" },
          "unit": { "name": "tbsp" },
          "amount": 1
        }
      ]
    }
  ]
}

Key Points for Agents:

• name and steps are required fields
• Each step must have instruction, order, and ingredients (can be empty array)
• Each ingredient must have food.name and amount
• Use null for unit when no unit applies (e.g., "to taste", "pinch")
• Common units: g, kg, ml, l, cup, tbsp, tsp, whole, pinch
• Times are in minutes
• Steps should be ordered sequentially (1, 2, 3, ...)

Destructive example — delete a recipe:

# Agent should first ask: "This will permanently delete recipe 42. This cannot be undone. Confirm deletion?"
# Only after explicit user confirmation:
tandoor delete 42 --force

Image example — upload an image to a recipe:

# Agent should first ask: "I will upload my-recipe-photo.jpg to recipe 42. Proceed?"
# Only after user confirms:
tandoor image 42 ./my-recipe-photo.jpg

Household example — create a household and generate an invite link:

# Agent should first ask: "I will create a new household named 'My Family'. This requires admin privileges. Proceed?"
# Only after user confirms:
tandoor household add "My Family"

# Agent should first ask: "I will create an invite link for household 1. This requires space-owner token. Proceed?"
# Only after user confirms:
tandoor household invite create 1 --email user@example.com --expires 2026-12-31

Shopping list example — add items and check them off:

# Agent should first ask: "I will add flour (500g) to the shopping list. Proceed?"
# Only after user confirms:
tandoor shopping add --food flour --amount 500 --unit g

# Agent should first ask: "This will mark ALL shopping items as checked. Confirm?"
# Only after explicit confirmation:
tandoor shopping check --all

Grocery category example — organise foods for a better shopping list:

# Read example — see what categories exist:
tandoor category list

# Read example — find foods with no category yet:
tandoor category uncategorised
tandoor category uncategorised --search pasta

# Agent should first ask: "I will assign 'milk' to the Dairy category. Proceed?"
# Only after user confirms:
tandoor category set milk --category Dairy
tandoor category set "olive oil" --category Oils
tandoor category set 42 --category "Canned Goods"   # by food ID

# Agent should first ask: "I will remove the category from 'milk'. Proceed?"
# Only after user confirms:
tandoor category set milk --unset

Cook log example — track when you cook a recipe:

# Agent should first ask: "I will add a cook log entry for recipe 42 (4 servings, rating 5). Proceed?"
# Only after user confirms:
tandoor cooklog add --recipe 42 --servings 4 --rating 5
tandoor cooklog add --recipe 42 --servings 4 --rating 5 --comment "Kids loved it"

# Read example — list cook logs for a specific recipe:
tandoor cooklog list --recipe 42 --format api

# Read example — find when you last had eggs:
tandoor cooklog ingredient eggs

# Read example — count how many times you had chicken last month:
tandoor cooklog ingredient chicken --startdate 2026-04-01 --enddate 2026-04-30

# Read example — find highly-rated egg recipes (4-5 stars):
tandoor cooklog ingredient eggs --min-rating 4

# Read example — find poorly-rated recipes from last month (1-2 stars):
tandoor cooklog list --startdate 2026-04-01 --enddate 2026-04-30 --max-rating 2

# Agent should first ask: "I will update cook log 2 with a new comment. Proceed?"
# Only after user confirms:
tandoor cooklog update 2 --comment "Next time add more garlic"

# Agent should first ask: "This will delete cook log entry 2. Confirm?"
# Only after explicit confirmation:
tandoor cooklog delete 2

Installation & Setup

Before first use:

1. Verify the npm package: https://www.npmjs.com/package/tandoor-cli
2. Review the source code: https://github.com/dcenatiempo/tandoor-cli
3. Recommended: Install the pinned version matching this skill: npm install -g tandoor-cli@<version> (see version in metadata above)
4. Generate a token with minimal required permissions (see SECURITY.md)
5. Consider using a test Tandoor instance first

Reference

• See references/SETUP.md for detailed setup documentation including authentication configuration
• See SECURITY.md for comprehensive security guidelines and token management best practices