Works with the AI Agents you already use
JSON Schema Conformance Gate for LLM Outputs
This skill checks one or many outputs against your schema and tells you exactly what is wrong and where.
$7
JSON Schema Conformance Gate for LLM Outputs
Example session with this skill installed
Our invoice extractor ran three times on the same PDF. Check run1.json, run2.json and run3.json against schema.json and tell me which ones are safe to load into billing.
- Read your context and instructions
- Compiled the json schema conformance
I ran json-schema-conform with schema.json on all three outputs. Exit code 1: two pass, one fails.
{
"schemaSha256": "e0be1a82f6ce5e9b05239495cb7b69aa40b9356f6658c02cdec713d35ce750d0",
"results": [
{ "file": "run1.json", "verdict": "PASS", "errors": [],
"canonicalSha256": "0e0c3f279536cdba516b7f0eaacbd4190abbdfcb265dc8023a94fa25e88a9d8d" },
{ "file": "run2.json", "verdict": "PASS", "errors": [],
"canonicalSha256": "0e0c3f279536cdba516b7f0eaacbd4190abbdfcb265dc8023a94fa25e88a9d8d" },
{ "file": "run3.json", "verdict": "FAIL",
"errors": [
{ "path": "$/invoice_id", "message": "string does not match pattern ^INV-[0-9]{5}$" },
{ "path": "$/total", "message": "type mismatch: expected number, got string" },
{ "path": "$/currency", "message": "value is not one of the allowed enum values" },
{ "path": "$/line_items[0]/qty", "message": "value 0 is below minimum=1" },
{ "path": "$/notes", "message": "unexpected property 'notes' (additionalProperties: false)" }
],
"canonicalSha256": "57493ece24a394bb73b3d735983f1cf81f5ef383e64b055272a4289fdae54bf8" }
],
"identicalContentGroups": [ [ "run1.json", "run2.json" ] ]
}
Summary
- run1 and run2 PASS and are the same content (same canonical hash), even though run2 has different key order and spacing. Load either one.
- run3 FAILS on five points: an ID that isn't INV-#####, the total sent as a string, "gbp" in lower case, a quantity of 0, and an extra "notes" field. Don't load it. I'd regenerate it rather than patch it by hand.
Connects securely to your tools. The creator never sees your data.
About this skill
Agent Trust: structured output that actually matches the contract
Models return JSON that looks right and then breaks your code: a number sent as a string, a lower-case enum, an extra field, a quantity of zero. This skill checks one or many outputs against your schema and tells you exactly what is wrong and where.
What it does
- Validates each JSON file against a JSON Schema and returns PASS or FAIL
- Lists every error with a JSON path (e.g.
$/line_items[0]/qty) and a plain message - Fingerprints each output with a canonical SHA-256 (sorted keys, normalised whitespace)
- Groups outputs whose content is identical even if formatting differs, which is useful for checking whether repeated runs drift
Supported keywords include type, required, properties, additionalProperties, patternProperties, enum, const, pattern, minimum/maximum (and exclusive), multipleOf, min/maxLength, min/maxItems, uniqueItems, items, allOf, anyOf, oneOf, not, local $ref and a best-effort format check.
When an agent should use it
- An agent or model returns JSON that downstream code will trust
- You compare several runs of the same prompt for structural drift
- You check a tool-call payload or config against its schema before use
What it outputs
JSON: the schema hash, per-file PASS/FAIL with error paths, a canonical SHA-256 per output, and groups of identical outputs. Exit 0 = all PASS; 1 = at least one FAIL.
Limitations
- Implements a practical subset of JSON Schema, not the full spec. Only local
#/...$refs are supported; remote refs are refused rather than fetched. formatchecks are best-effort pattern checks.- It validates; it does not repair. Fixing or regenerating the output is up to the agent.
Open source, packaged skill
Original work by Simon Gant. The core logic is public under the MIT licence (kar-conform). This listing is the packaged, tested agent skill: SKILL.md workflow, CLI, receipt output and docs. Requires Node.js 18+, zero dependencies, no network, no API keys.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed today
- Software
Node.js 18+
Permissions
Runs one local Node.js script (`node scripts/json-schema-conform.js`) through the agent's shell tool. Reads only the schema and output files you pass it. No network access (remote `$ref`s are refused, not fetched), no file writes, no environment variables read. Pure compute.