structured-output

Included with Lifetime

$97 forever

Get reliable typed JSON from LLMs using constrained decoding, JSON Schema, Zod, Pydantic, and Instructor. Use when implementing structured output, JSON schema validation, typed API responses, constrained generation, Zod schemas, Pydantic models, schema design for LLM extraction, streaming structured data, or debugging malformed JSON responses from models.

Design

What this skill does

# Structured Output

Structured output guarantees valid typed JSON from LLMs — the model can only emit tokens that conform to your schema. The question isn't how it works but when you need it.

**Use when:** data extraction, API responses, agent tool outputs, any downstream code that parses the response.

**Don't use when:** prose, creative writing, chat, or when the quality cost outweighs the reliability gain.

For code examples and provider-specific patterns, see [reference.md](reference.md).

## The Quality Problem

Constrained decoding has a cost. Tam et al. (2024) showed JSON constraints dropped math accuracy by 26–32 percentage points across tested models. The mechanism: forced key ordering means the model must generate answer tokens before reasoning tokens, disrupting sequential reasoning.

The fix is schema design, not giving up on structured output. Enforcement is syntactic, not semantic — valid JSON can still be semantically wrong.

## Schema Design: The #1 Lever

The PARSE paper (2025) found structural reorganization was the top optimization for extraction accuracy — more impactful than model choice or prompt engineering. A single ambiguous field name (`final_choice` alongside `potential_final_choices`) collapsed accuracy from **95% → 4.5%** in Instructor's experiments.

### The reasoning field trick

Put a `reasoning` or `thinking` field **first** in your schema. LLMs generate keys in order — reasoning before answer tokens restores chain-of-thought within structured output. Instructor showed this alone recovers accuracy from **33% → 92%** on GSM8K.

```json
{
"type": "object",
"properties": {
"reasoning": { "type": "string", "description": "Step-by-step analysis before concluding" },
"answer": { "type": "string" },
"confidence": { "type": "string", "enum": ["high", "medium", "low"] }
},
"required": ["reasoning", "answer", "confidence"],
"additionalProperties": false
}
```

### Rules that matter

- Flat over nested. PARSE found flattening was the most frequent successful optimization (55% of cases).
- Specific field descriptions: "Full legal name including first and last" beats "The name". Specific descriptions reduced errors by up to 64.7% (PARSE).
- String enums for constrained choices — reduces downstream cleaning by ~70% (PARSE).
- `minimum`/`maximum`/`pattern` — neither Anthropic nor OpenAI enforces these at the grammar level. Validate post-generation.
- Max 24 optional parameters (Anthropic hard limit). Each optional field roughly doubles grammar state space.
- Recursive schemas not supported on Anthropic.

These patterns apply directly to [tool-design](../tool-design/SKILL.md) — tool schemas are structured output schemas.

## The Production Loop

```
Type system (Zod/Pydantic)
→ JSON Schema
→ LLM with constrained decoding
→ Validate (syntactic: already guaranteed; semantic: your validators)
→ Retry with error feedback on failure
```

Libraries doing this loop:
- **Python**: Instructor wraps OpenAI/Anthropic/Groq/etc, feeds validation errors back into the conversation, retries up to `max_retries`.
- **TypeScript/OpenAI**: `zodResponseFormat` + `beta.chat.completions.parse()` — returns `.parsed` typed directly.
- **TypeScript/Anthropic**: `zodOutputFormat` helper from `@anthropic-ai/sdk/helpers/zod`.
- **TypeScript/any provider**: Vercel AI SDK `generateObject` / `streamObject` with a Zod schema.

Zod v4 has native `z.toJSONSchema()` for converting schemas to JSON Schema format.

---

For mechanism details, provider comparison, streaming patterns, and failure modes, see [reference.md](reference.md).

Files: 2

Size: 19.7 KB

Complexity: 27/100

Category: Design

Source: https://github.com/captaincrouton89/crouton-kit/tree/main/plugins/authoring/skills/structured-output

Related in Design

contribute

Included

Local-only OSS contribution command center. Auto-refreshes the user's in-flight PR and issue state on invoke so conversations start with full context — no need to brief Claude on what's in flight. Helps the user find issues to contribute to on GitHub, builds per-repo dossiers of what each upstream expects (CLA, DCO, branch convention, AI policy, draft-first, review bots, issue templates), runs deterministic gates before any external action so AI-assisted contributions don't reach maintainers as slop. State is markdown-only: candidate files at ~/.contribute-system/candidates/, repo dossiers at ~/.contribute-system/research/, append-only event log at ~/.contribute-system/log.jsonl. No database, no cloud calls. Use when the user asks about their PRs / issues / contributions, wants to find new work to take on, claim an issue, build/refresh a repo's dossier, or draft a Design Issue or PR. Trigger with "/contribute", "what's my PR status", "find a contribution", "claim issue X", "draft a Design Issue for Y", "refresh dossier for Z".

Designscripts

architectural-analysis

Included

User-triggered deep architectural analysis of a codebase or scoped subtree across eight modes — information architecture, data flow, integration points, UI surfaces, interaction patterns, data model, control flow, and failure modes. This skill should be used when the user asks to "diagram this codebase," "map the architecture," "show the data flow," "give me an ERD," "trace control flow," "find the integration points," "verify the layout pattern," "audit the UX architecture," or any similar request whose primary deliverable is mermaid diagrams plus cited reports under docs/architecture/. Dispatches haiku/sonnet sub-agents in parallel for per-mode exploration, then verifies every citation mechanically before any node lands in a diagram. Not for one-off prose explanations of code (use code-explanation) or for high-level system design from scratch (use system-design).

Designscripts

mcp

Included

Model Context Protocol (MCP) server development and tool management. Languages: Python, TypeScript. Capabilities: build MCP servers, integrate external APIs, discover/execute MCP tools, manage multi-server configs, design agent-centric tools. Actions: create, build, integrate, discover, execute, configure MCP servers/tools. Keywords: MCP, Model Context Protocol, MCP server, MCP tool, stdio transport, SSE transport, tool discovery, resource provider, prompt template, external API integration, Gemini CLI MCP, Claude MCP, agent tools, tool execution, server config. Use when: building MCP servers, integrating external APIs as MCP tools, discovering available MCP tools, executing MCP capabilities, configuring multi-server setups, designing tools for AI agents.

Designscripts

react-native-skia

Included

Design, build, debug, and optimise high-polish animated graphics in React Native or Expo using @shopify/react-native-skia, Reanimated, and Gesture Handler. Use when the user wants canvas-driven UI, shaders, paths, rich text, image filters, sprite fields, Skottie, video frames, snapshots, web CanvasKit setup, or performance tuning for custom motion-heavy elements such as loaders, hero art, cards, charts, progress indicators, particle systems, or gesture-driven surfaces. Also use when the user asks for fluid, glow, glass, blob, parallax, 60fps/120fps, or GPU-friendly animated effects in React Native, even if they do not explicitly say "Skia". Do not use for ordinary form/layout work with standard views.

Designscripts

plaid

Included

Product Led AI Development — guides founders from idea to launched product. Six capabilities: Idea (discover a product idea), Validate (pressure-test the idea against fatal flaws, problem reality, competition, and 2-week MVP feasibility), Plan (vision intake + document generation), Design (translate image references into a design.md spec), Launch (go-to-market strategy), and Build (roadmap execution). Use when someone says "PLAID", "plaid idea", "help me find an idea", "product idea", "idea from my business", "idea from my expertise", "plaid validate", "validate my idea", "pressure-test", "is this idea good", "find fatal flaws", "validate the problem", "plan a product", "define my vision", "generate a PRD", "product strategy", "plaid design", "design from image", "translate image to design", "create design.md", "extract design tokens", "plaid launch", "go-to-market", "launch plan", "GTM strategy", "launch playbook", "plaid build", "build the app", "start building", or "execute the roadmap".

Designscripts

nextjs-framer-motion-animations

Included

Adds production-safe Motion for React or Framer Motion animations to Next.js apps, including reveal, hover and tap micro-interactions, whileInView, stagger, AnimatePresence, layout and layoutId transitions, reorder, scroll-linked UI, and lightweight route-content transitions. Use when the user asks to add, refactor, or debug Motion or Framer Motion in App Router or Pages Router codebases, especially around server/client boundaries, reduced motion, LazyMotion, bundle size, hydration, or route transitions. Avoid for GSAP-style timelines, WebGL or 3D scenes, heavy scroll storytelling, or CSS-only effects unless Motion is explicitly requested.

Designscripts