neuron-analyze-api

name: neuron-analyze-api description: Analyze arbitrary-format API lists and extract structured understanding for neuron-ui page generation. Use when users provide API documentation in any format (Swagger/OpenAPI, Postman collections, cURL commands, text descriptions, tables, screenshots) and need it analyzed before page generation. Triggers on requests like "analyze this API", "understand these endpoints", "what pages can I build from this API", or when preparing API context for the neuron-generate-page skill.

Neuron Analyze API

Parse and understand API documentation in any format, producing a structured analysis that feeds into page generation.

Workflow

1. Detect input format   → Swagger? Postman? Text? Table? cURL?
2. Extract resources     → Group endpoints by resource entity
3. Extract endpoints     → Method, path, summary per endpoint
4. Extract fields        → Field name, type, constraints per resource
5. Infer field types     → Map to neuron field type system
6. Detect API patterns   → CRUD? Stats? Auth? Custom?
7. Output structured analysis

Input Format Detection

No parsing required — just understand the content. Supported inputs:

Field Type Inference

Map discovered fields to neuron's type system:

From explicit schema (Swagger/Postman)

type: string               → string
type: string + format:date → date
type: string + enum:[...]  → string:enum
type: integer/number       → number
type: boolean              → boolean
type: array + items:string → array:string
type: array + items:object → array:object
type: object               → object:nested

From field names (text/informal descriptions)

Read references/field-inference-rules.md for complete inference table.

Quick reference:

• *_at, *_date, created_at → date/datetime
• avatar, image, cover → string:image
• status, type, role → string:enum (find possible values from context)
• description, content, body → string:long
• tags, labels → array:string
• is_*, has_* → boolean
• *_id (referencing another resource) → object:user or object:nested

API Pattern Detection

After extracting all endpoints for a resource, detect the pattern:

CRUD pattern:
  GET  /resources       (list)     ─┐
  GET  /resources/:id   (detail)    ├── All present → CRUD
  POST /resources       (create)    │
  PUT  /resources/:id   (update)    │
  DELETE /resources/:id (delete)   ─┘

Stats/Dashboard pattern:
  GET /stats            ─┐
  GET /analytics         ├── Multiple read-only aggregate endpoints → Dashboard
  GET /summary          ─┘

Auth pattern:
  POST /login           ─┐
  POST /register         ├── Auth-related endpoints → Auth form
  POST /verify-otp      ─┘

Partial CRUD:
  Missing DELETE → No delete confirm
  Missing PUT   → No edit form (read-only detail)
  Only GET list → Read-only list page

Output Format

Produce a structured analysis:

{
  "resources": [
    {
      "name": "competition",
      "displayName": "赛事",
      "endpoints": [
        { "method": "GET", "path": "/api/competitions", "type": "list", "queryParams": ["status", "keyword", "page"] },
        { "method": "POST", "path": "/api/competitions", "type": "create" },
        { "method": "GET", "path": "/api/competitions/:id", "type": "detail" },
        { "method": "PUT", "path": "/api/competitions/:id", "type": "update" },
        { "method": "DELETE", "path": "/api/competitions/:id", "type": "delete" }
      ],
      "fields": [
        { "name": "name", "type": "string", "required": true },
        { "name": "description", "type": "string:long" },
        { "name": "status", "type": "string:enum", "options": ["draft", "active", "ended"] },
        { "name": "cover", "type": "string:image" },
        { "name": "prize", "type": "number" },
        { "name": "start_date", "type": "date" },
        { "name": "tags", "type": "array:string" },
        { "name": "created_by", "type": "object:user" }
      ],
      "pattern": "crud"
    }
  ]
}

This output is consumed by the neuron-generate-page skill to generate Page Schema.

Resources

references/

• field-inference-rules.md — Complete field name pattern → neuron type inference rules with examples.