How do I install the Structured Output skill?

Run npx skillmds@latest add xspoonai/structured-output in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Structured Output safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Structured Output?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Structured Output free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Structured Output?

XSpoonAi (@xspoonai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Structured Output by xspoonai — Agent Skill

Name: Structured Output
Author: XSpoonAi

Structured Output

Stable structured JSON output for agents using strict schemas. Use when you need deterministic JSON responses, schema validation, or streaming JSON object assembly (generateObject/streamObject-style) with retry-and-repair loops.

XSpoonAi dd73692 6 files · 13.4 KB Updated Sep 17, 2026

File contents

Structured Output (Strict JSON)

Implement stable, schema-validated JSON outputs for agents. Use this when responses must be valid JSON, match a schema exactly, and be reliable under retries/streaming.

Core Workflow

Define a strict schema (JSON Schema or Pydantic).
Generate JSON-only output (no prose).
Validate output against schema.
If invalid, run a repair pass with a minimal error-focused prompt.
For streaming, emit incremental JSON with a deterministic assembly buffer.

Prompt Pattern (Non-Streaming)

Use a strict system instruction and a minimal user prompt. Always require JSON only. Input can be empty; output must still match the schema.

Template:

SYSTEM:
You are a JSON generator. Output ONLY valid JSON that matches the given schema.
Do not include Markdown, comments, or trailing text.
If a field is unknown, use null (unless schema disallows it).

USER:
Return JSON that matches this schema:
<SCHEMA>

Input:
<INPUT>

Prompt Pattern (Streaming)

Use the same schema but ensure the model emits a single JSON object and nothing else. Stream tokens and buffer until a valid JSON parse succeeds.

Implementation References

JSON Schema + retry/repair loop: see references/json_schema_patterns.md
Pydantic models + parsing: see references/pydantic_patterns.md
Python examples: scripts/generate_object.py, scripts/stream_object.py

Guardrails

Reject non-JSON output immediately; re-prompt with error-only instructions.
Disallow extra keys unless additionalProperties is explicitly true.
Keep schemas small and explicit; prefer enums for constrained values.
Log schema + error details for debugging (do not expose stack traces to end users).

---name: structured-outputdescription: Stable structured JSON output for agents using strict schemas. Use when you need deterministic JSON responses, schema validation, or streaming JSON object assembly (generateObject/streamObject-style) with retry-and-repair loops.---# Structured Output (Strict JSON)Implement stable, schema-validated JSON outputs for agents. Use this when responses must be valid JSON, match a schema exactly, and be reliable under retries/streaming.## Core Workflow1. Define a strict schema (JSON Schema or Pydantic).2. Generate JSON-only output (no prose).3. Validate output against schema.4. If invalid, run a repair pass with a minimal error-focused prompt.5. For streaming, emit incremental JSON with a deterministic assembly buffer.## Prompt Pattern (Non-Streaming)Use a strict system instruction and a minimal user prompt. Always require JSON only. Input can be empty; output must still match the schema.Template:```SYSTEM:You are a JSON generator. Output ONLY valid JSON that matches the given schema.Do not include Markdown, comments, or trailing text.If a field is unknown, use null (unless schema disallows it).USER:Return JSON that matches this schema:<SCHEMA>Input:<INPUT>```## Prompt Pattern (Streaming)Use the same schema but ensure the model emits a single JSON object and nothing else. Stream tokens and buffer until a valid JSON parse succeeds.## Implementation References- JSON Schema + retry/repair loop: see `references/json_schema_patterns.md`- Pydantic models + parsing: see `references/pydantic_patterns.md`- Python examples: `scripts/generate_object.py`, `scripts/stream_object.py`## Guardrails- Reject non-JSON output immediately; re-prompt with error-only instructions.- Disallow extra keys unless `additionalProperties` is explicitly true.- Keep schemas small and explicit; prefer enums for constrained values.- Log schema + error details for debugging (do not expose stack traces to end users).

Frequently asked questions