Prompt Variable Validator

Overview

You send a prompt template and the variables you plan to render it with, and we tell you whether the render will come out complete. We find every placeholder in one of four supported syntaxes, compare the names against your variables, and report what is missing, what you sent but never used, and what looks like a placeholder but is not one we will interpret — a section tag, a filter or an expression comes back named rather than quietly ignored, because pretending not to see it would tell you the template is fully covered when it is not. Nothing here calls a model, renders the template or reads your variable values beyond checking whether they are empty. Use it when you want the "you are missing customer_name" answer before the call, not a prompt with an unfilled hole in it afterwards.

A prompt with an unfilled hole in it does not fail loudly — it goes to the model, costs you a call, and comes back with an answer that quietly names a placeholder instead of a customer. Checking coverage first turns that into a cheap, local, list-shaped answer you can act on.

When to use it

Scope

Non-goals

• Does not render the template — you get a verdict, never filled-in text
• Does not call a model, an embedding endpoint or any inference API
• Does not decide whether a template is allowed to reference something — communication.template_validate answers that against an allowlist
• Does not interpret sections, conditionals, loops or filters; it names them instead
• Does not resolve dotted names into nested objects: customer.name looks for a variable literally called customer.name

Properties

Capabilities

Cost & billing

Flat rate per invocation

File handling

How we handle your data

Input schema

Output schema

API contract

curl -sS -X POST "$BASE/v1/execute" -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"skillKey":"ai.prompt_variable_validate","input":{"promptTemplate":"You are a support agent for {{company}}. Write a reply to {{customer_name}} about order {{order_id}}.","variables":{"company":"Northwind Supplies","customer_name":"Priya Raman","order_id":"NW-84120"}}}'

import { Squillify } from "@squillify/sdk";

const sq = new Squillify({ apiKey: process.env.SQUILLIFY_API_KEY });

const { output } = await sq.execute("ai.prompt_variable_validate", {"promptTemplate":"You are a support agent for {{company}}. Write a reply to {{customer_name}} about order {{order_id}}, and sign it as {{company}}.","variables":{"company":"Northwind Supplies","customer_name":"Priya Raman","order_id":"NW-84120"}});

Errors

Other limits

Limitations

• Does not render the template — you get a verdict, never filled-in text
• Does not call a model, an embedding endpoint or any inference API
• Does not decide whether a template is allowed to reference something — communication.template_validate answers that against an allowlist
• Does not interpret sections, conditionals, loops or filters; it names them instead
• Does not resolve dotted names into nested objects: customer.name looks for a variable literally called customer.name
• Rendering the finished prompt, which this never does
• Deciding whether a template is permitted to reference sensitive fields
• Templates written in a full template language with loops and conditionals you expect to be evaluated

How to call

POST /v1/execute (requires authentication)

{
  "skillKey": "ai.prompt_variable_validate",
  "input": {
    "promptTemplate": "You are a support agent for {{company}}. Write a reply to {{customer_name}} about order {{order_id}}, and sign it as {{company}}.",
    "variables": {
      "company": "Northwind Supplies",
      "customer_name": "Priya Raman",
      "order_id": "NW-84120"
    }
  }
}

Examples

{
  "promptTemplate": "You are a support agent for {{company}}. Write a reply to {{customer_name}} about order {{order_id}}, and sign it as {{company}}.",
  "variables": {
    "company": "Northwind Supplies",
    "customer_name": "Priya Raman",
    "order_id": "NW-84120"
  }
}

{
  "placeholders": [
    {
      "name": "company",
      "occurrences": 2,
      "firstSpan": {
        "start": 28,
        "end": 39
      }
    },
    {
      "name": "customer_name",
      "occurrences": 1,
      "firstSpan": {
        "start": 58,
        "end": 75
      }
    },
    {
      "name": "order_id",
      "occurrences": 1,
      "firstSpan": {
        "start": 88,
        "end": 100
      }
    }
  ],
  "valid": true,
  "blank": [],
  "summary": {
    "placeholderCount": 4,
    "distinctPlaceholderCount": 3,
    "variableCount": 3,
    "missingCount": 0,
    "unusedCount": 0,
    "syntax": "MUSTACHE",
    "truncated": false
  },
  "missing": [],
  "unused": [],
  "unsupportedForms": []
}

{
  "promptTemplate": "Dear {{customer_name}}, your invoice {{invoice_id}} is due on {{due_date}}.{{#if overdue}} It is overdue.{{/if}}",
  "variables": {
    "customer_name": "Priya Raman",
    "invoice_id": "  ",
    "locale": "en-GB"
  },
  "requiredVariables": [
    "tenant_id"
  ],
  "treatBlankAsMissing": true
}

{
  "placeholders": [
    {
      "name": "customer_name",
      "occurrences": 1,
      "firstSpan": {
        "start": 5,
        "end": 22
      }
    },
    {
      "name": "invoice_id",
      "occurrences": 1,
      "firstSpan": {
        "start": 37,
        "end": 51
      }
    },
    {
      "name": "due_date",
      "occurrences": 1,
      "firstSpan": {
        "start": 62,
        "end": 74
      }
    }
  ],
  "valid": false,
  "blank": [
    "invoice_id"
  ],
  "summary": {
    "placeholderCount": 3,
    "distinctPlaceholderCount": 3,
    "variableCount": 3,
    "missingCount": 2,
    "unusedCount": 1,
    "syntax": "MUSTACHE",
    "truncated": false
  },
  "missing": [
    "due_date",
    "tenant_id"
  ],
  "unused": [
    "locale"
  ],
  "unsupportedForms": [
    {
      "form": "{{#if overdue}}",
      "reason": "CONTROL_TAG",
      "span": {
        "start": 75,
        "end": 90
      }
    },
    {
      "form": "{{/if}}",
      "reason": "CONTROL_TAG",
      "span": {
        "start": 105,
        "end": 112
      }
    }
  ]
}

Recommended use cases

• Catch a missing variable before you spend a model call rendering a broken prompt
• Show an operator exactly which fields a prompt template still needs
• Fail a deployment when a template picked up a control tag your renderer cannot fill

Avoid for

• Rendering the finished prompt, which this never does
• Deciding whether a template is permitted to reference sensitive fields
• Templates written in a full template language with loops and conditionals you expect to be evaluated

Related skills

Composable with

Safety & privacy

• Runs without calling any external service.
• Your payloads are never stored permanently.
• Telemetry captures metadata only — never your payload contents.
• We never train models on your payloads.
• No human reviews your payloads.

Tags