Tool Schema Minifier

Overview

A tool schema is re-sent on every agent turn, so every annotation in it is paid for again and again. We remove the parts that cannot change whether a document validates — titles, descriptions, comments, examples, definitions nothing references, keywords set to the value the specification already assumes, and a type a sibling const or single-member enum already forces — and we publish the keep list too, so you can see that type, required, bounds, patterns, formats, default, deprecated, readOnly and writeOnly all survive. Each removal comes back with its pointer and reason, and the saving is reported in tokens from a named tokenizer with its pinned engine version, measured over the same whitespace-free serialisation on both sides.

A tool schema is not paid for once at design time; it is paid for on every agent turn, in the context window, for as long as the tool exists. Half of a typical one is prose a validator never reads.

When to use it

Scope

Non-goals

• We never remove a keyword that can change a validation verdict
• We never remove default, deprecated, readOnly or writeOnly — they change behaviour, not size
• We never reorder keys, rename anything, or rewrite a schema into another shape
• We never fetch a reference: anything outside the document you sent is refused

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":"schema.minify_tool_schema","input":{"schema":{"type":"object","title":"Order","properties":{"id":{"type":"string"}}}}}'

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

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

const { output } = await sq.execute("schema.minify_tool_schema", {"tokenizer":"cl100k_base","schema":{"title":"Order","type":"object","properties":{"id":{"type":"string","description":"The order identifier.","minLength":0}}}});

Errors

Other limits

Limitations

• We never remove a keyword that can change a validation verdict
• We never remove default, deprecated, readOnly or writeOnly — they change behaviour, not size
• We never reorder keys, rename anything, or rewrite a schema into another shape
• We never fetch a reference: anything outside the document you sent is refused
• Producing a schema for humans to read
• Removing sample values that carry secrets, which schema.redact_schema_samples does

How to call

POST /v1/execute (requires authentication)

{
  "skillKey": "schema.minify_tool_schema",
  "input": {
    "tokenizer": "cl100k_base",
    "schema": {
      "title": "Order",
      "type": "object",
      "properties": {
        "id": {
          "type": "string",
          "description": "The order identifier.",
          "minLength": 0
        }
      }
    }
  }
}

Examples

{
  "tokenizer": "cl100k_base",
  "schema": {
    "title": "Order",
    "type": "object",
    "properties": {
      "id": {
        "type": "string",
        "description": "The order identifier.",
        "minLength": 0
      }
    }
  }
}

{
  "summary": {
    "removalCount": 3,
    "originalBytes": 123,
    "minifiedBytes": 55,
    "nodeCount": 8,
    "preserved": [],
    "truncatedDescriptions": 0,
    "truncated": false
  },
  "removals": [
    {
      "pointer": "/title",
      "keyword": "title",
      "reason": "ANNOTATION",
      "action": "REMOVED"
    },
    {
      "pointer": "/properties/id/description",
      "keyword": "description",
      "reason": "ANNOTATION",
      "action": "REMOVED"
    },
    {
      "pointer": "/properties/id/minLength",
      "keyword": "minLength",
      "reason": "REDUNDANT_DEFAULT_VALUE",
      "action": "REMOVED"
    }
  ],
  "tokenEstimate": {
    "originalTokens": 28,
    "minifiedTokens": 14,
    "tokensSaved": 14,
    "tokenizer": {
      "id": "cl100k_base",
      "family": "byte-level-bpe",
      "engineId": "jtokkit",
      "libraryVersion": "jtokkit-1.1.0"
    },
    "serialization": "CANONICAL_MINIFIED"
  },
  "minifiedSchema": {
    "type": "object",
    "properties": {
      "id": {
        "type": "string"
      }
    }
  }
}

{
  "schema": {
    "type": "object",
    "description": "An order.",
    "properties": {
      "id": {
        "type": "string",
        "description": "The order identifier."
      }
    },
    "$defs": {
      "unused": {
        "type": "integer"
      }
    }
  },
  "reportRemovals": false,
  "preserve": [
    "DESCRIPTIONS"
  ]
}

{
  "summary": {
    "removalCount": 2,
    "originalBytes": 157,
    "minifiedBytes": 119,
    "nodeCount": 10,
    "preserved": [
      "DESCRIPTIONS"
    ],
    "truncatedDescriptions": 0,
    "truncated": false
  },
  "tokenEstimate": {
    "originalTokens": 36,
    "minifiedTokens": 25,
    "tokensSaved": 11,
    "tokenizer": {
      "id": "cl100k_base",
      "family": "byte-level-bpe",
      "engineId": "jtokkit",
      "libraryVersion": "jtokkit-1.1.0"
    },
    "serialization": "CANONICAL_MINIFIED"
  },
  "minifiedSchema": {
    "type": "object",
    "description": "An order.",
    "properties": {
      "id": {
        "type": "string",
        "description": "The order identifier."
      }
    }
  }
}

Recommended use cases

• Shrink a tool schema before publishing it to agents that pay for it every turn
• Measure how much of a schema's size is documentation rather than constraint
• Strip a schema for a machine consumer while keeping the annotated copy in Git

Avoid for

• Producing a schema for humans to read
• Removing sample values that carry secrets, which schema.redact_schema_samples does

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