Agent-Skills.md

Agent Skills: SAP CAP-Capire Development Skill

|

UncategorizedID: secondsky/sap-skills/sap-cap-capire

Repository

secondsky/sap-skills
secondskyLicense: GPL-3.0
18243

Install this agent skill to your local

pnpm dlx add-skill https://github.com/secondsky/sap-skills/tree/HEAD/plugins/sap-cap-capire/skills/sap-cap-capire

Skill Files

Browse the full folder contents for sap-cap-capire.

Download Skill

Loading file tree…

plugins/sap-cap-capire/skills/sap-cap-capire/SKILL.md

Skill Metadata

Name
sap-cap-capire
Description
|

SAP CAP-Capire Development Skill

Related Skills

  • sap-fiori-tools: Use for UI layer development, Fiori Elements integration, and frontend application generation
  • sapui5: Use for custom UI development, advanced UI patterns, and freestyle application building
  • sap-btp-cloud-platform: Use for deployment options, Cloud Foundry/Kyma configuration, and BTP service integration
  • sap-hana-cli: Use for database management, schema inspection, and HDI container administration
  • sap-abap: Use for ABAP system integration, external service consumption, and SAP extensions
  • sap-btp-best-practices: Use for production deployment patterns and architectural guidance
  • sap-ai-core: Use when adding AI capabilities to CAP applications or integrating with SAP AI services
  • sap-cloud-sdk-ai: Use for SDK-level AI integration (chat completion, streaming, tool calling) in CAP event handlers
  • sap-cloud-sdk-ai-python: Use for Python-based AI integration with CAP Java or standalone BTP services
  • sap-api-style: Use when documenting CAP OData services or following API documentation standards
  • sap-dependency-security: Use for secure dependency, lockfile, supply-chain, and exact MCP server pin controls in CAP service repos

When to Use This Skill

Use this skill when creating CAP projects, modeling CDS entities/services, implementing Node.js or Java event handlers, configuring HANA/SQLite/PostgreSQL persistence, deploying to BTP Cloud Foundry or Kyma, adding Fiori UIs, configuring authorization/multitenancy/messaging, or using CAP MCP/LSP tooling.

Common Issues

| Issue | First check | |-------|-------------| | cds watch or cds serve fails | Verify @sap/cds-dk, Node.js version, and project package.json scripts. | | Entity/service not found | Use CAP MCP search_model when available, then inspect db/ and srv/ CDS files. | | HANA deployment fails | Check HDI service binding, mta.yaml, and the HANA deployment references. | | Authorization behaves unexpectedly | Review @requires, @restrict, XSUAA/IAS bindings, and user role mappings. |

Table of Contents

Quick Start

Project Initialization

# Use an approved CAP toolchain:
# - project-local devDependencies
# - user-local npm prefix
# - enterprise-managed Node/CAP installation
# Ensure cds and optional cds-lsp commands are on PATH.
cds --version

# Create new project
cds init <project-name>
cds init <project-name> --add sample,hana

# Start development server with live reload
cds watch

# Add capabilities
cds add hana          # SAP HANA database
cds add sqlite        # SQLite for development
cds add xsuaa         # Authentication
cds add mta           # Cloud Foundry deployment
cds add multitenancy  # SaaS multitenancy
cds add typescript    # TypeScript support

Basic Entity Example

using { cuid, managed } from '@sap/cds/common';

namespace my.bookshop;

entity Books : cuid, managed {
  title       : String(111) not null;
  author      : Association to Authors;
  stock       : Integer;
  price       : Decimal(9,2);
}

entity Authors : cuid, managed {
  name        : String(111);
  books       : Association to many Books on books.author = $self;
}

Basic Service

using { my.bookshop as my } from '../db/schema';

service CatalogService @(path: '/browse') {
  @readonly entity Books as projection on my.Books;
  @readonly entity Authors as projection on my.Authors;
  
  @requires: 'authenticated-user'
  action submitOrder(book: Books:ID, quantity: Integer) returns String;
}

MCP Integration

This skill integrates with the official CAP MCP (Model Context Protocol) server, providing AI agents with live access to your project's compiled CDS model and CAP documentation.

Available MCP Tools:

  • search_model - Fuzzy search for CDS entities, services, actions, and relationships in your compiled CSN model
  • search_docs - Semantic search through CAP documentation for syntax, patterns, and best practices

Key Benefits:

  • Instant Model Discovery: Query your project's entities, associations, and services without reading files
  • Context-Aware Documentation: Find relevant CAP documentation based on semantic similarity, not keywords
  • Zero Configuration: No credentials or environment variables required
  • Offline-Capable: All searches are local (model) or cached (docs)

Setup: See MCP Integration Guide for configuration with Claude Code, opencode, or GitHub Copilot. MCP package pins are governed by sap-dependency-security and validated by npm run validate:mcp-security.

Use Cases: See MCP Use Cases for illustrative local workflow examples and planning assumptions, not repository-verified ROI.

Agent Integration: The specialized agents (cap-cds-modeler, cap-service-developer, cap-project-architect, cap-performance-debugger) automatically use these MCP tools as part of their workflows.

CAP MCP and LSP Routing

Use MCP first for local model and docs questions, then fall back to direct file search when MCP is unavailable. Use rg -n "<entity|service|aspect|annotation|handler|cds compile|deployment>" references/*.md srv db app to locate the narrowest reference before loading long CAP guides.

  • Use references/mcp-integration.md for MCP configuration and package pin checks.
  • Use references/mcp-use-cases.md only for workflow selection and illustrative impact examples.
  • Use .lsp.json as a Claude-compatible sidecar for CAP editor integration; other harnesses should not assume it is auto-loaded.
  • For Codex, OpenCode, editors, or other LSP-capable clients, configure the command manually as node <sap-cap-capire-plugin-root>/lsp/cds-lsp-launcher.mjs --stdio. This still requires @sap/cds-lsp to be installed through an approved project-local devDependency, user-local npm prefix, or enterprise-managed toolchain, with cds-lsp available on PATH.
  • Without LSP integration, use the Markdown guidance, bundled references, rg, and CAP CLI checks directly.
  • Mark live deployment, HANA, XSUAA, and multitenancy verification pending unless the target project or tenant evidence is available.

Project Structure

project/
├── app/              # UI content (Fiori, UI5)
├── srv/              # Service definitions (.cds, .js/.ts)
├── db/               # Data models and schema
│   ├── schema.cds    # Entity definitions
│   └── data/         # CSV seed data
├── package.json      # Dependencies and CDS config
└── .cdsrc.json       # CDS configuration (optional)

Core Concepts

CDS Built-in Types

| CDS Type | SQL Mapping | Common Use | |----------|-------------|------------| | UUID | NVARCHAR(36) | Primary keys | | String(n) | NVARCHAR(n) | Text fields | | Integer | INTEGER | Whole numbers | | Decimal(p,s) | DECIMAL(p,s) | Monetary values | | Boolean | BOOLEAN | True/false | | Date | DATE | Calendar dates | | Timestamp | TIMESTAMP | Date/time |

Common Aspects

using { cuid, managed, temporal } from '@sap/cds/common';
// cuid = UUID key
// managed = createdAt, createdBy, modifiedAt, modifiedBy
// temporal = validFrom, validTo

Event Handlers (Node.js)

// srv/cat-service.js
module.exports = class CatalogService extends cds.ApplicationService {
  init() {
    const { Books } = this.entities;
    
    // Before handlers - validation
    this.before('CREATE', Books, req => {
      if (!req.data.title) req.error(400, 'Title required');
    });
    
    // On handlers - custom logic
    this.on('submitOrder', async req => {
      const { book, quantity } = req.data;
      // Custom business logic
      return { success: true };
    });
    
    return super.init();
  }
}

Basic CQL Queries

const { Books } = cds.entities;

// SELECT with conditions
const books = await SELECT.from(Books)
  .where({ stock: { '>': 0 } })
  .orderBy('title');

// INSERT
await INSERT.into(Books)
  .entries({ title: 'New Book', stock: 10 });

// UPDATE
await UPDATE(Books, bookId)
  .set({ stock: { '-=': 1 } });

AI Integration

CAP applications integrate with SAP AI Core via the SAP Cloud SDK for AI. The recommended pattern uses the Orchestration Service through CAP event handlers, with all credential management handled by BTP service bindings.

Service Binding (MTA)

resources:
  - name: my-ai-core
    type: org.cloudfoundry.managed-service
    parameters:
      service: aicore
      service-plan: extended

Local Development (Hybrid Mode)

cds bind -2 <AICORE_INSTANCE> && cds-tsx watch --profile hybrid

Event Handler Pattern (Node.js/TypeScript)

import { OrchestrationClient } from '@sap-ai-sdk/orchestration';

module.exports = class AnalysisService extends cds.ApplicationService {
  async init() {
    const { Feedback } = this.entities;

    this.on('analyzeFeedback', async (req) => {
      const userText = req.data.text;

      const client = new OrchestrationClient({
        promptTemplating: {
          model: { name: 'gpt-4o' },
          prompt: [
            { role: 'system', content: 'Categorize feedback as JSON: sentiment, category, urgency.' },
            { role: 'user', content: '{{?userText}}' }
          ]
        }
      });

      const response = await client.chatCompletion({
        placeholderValues: { userText }
      });

      const aiResult = response.getContent();

      await INSERT.into('FeedbackResults').entries({
        originalText: userText,
        analysisJson: aiResult
      });

      return aiResult;
    });

    return super.init();
  }
};

Asynchronous Processing (Production Pattern)

LLM calls can take 30-60 seconds. Never process them synchronously in production — the BTP load balancer will timeout before the LLM responds.

this.on('analyzeFeedback', async (req) => {
  const id = await INSERT.into('FeedbackResults').entries({
    originalText: req.data.text,
    status: 'processing'
  });

  cds.spawn(() => processWithLLM(id, req.data.text));

  return req.reply(202, { id, status: 'processing' });
});

async function processWithLLM(id, text) {
  const response = await client.chatCompletion({
    placeholderValues: { userText: text }
  });
  await UPDATE('FeedbackResults', id).set({
    analysisJson: response.getContent(),
    status: 'completed'
  });
}

HANA Vector Type for RAG

entity Documents {
  key id    : UUID;
  content   : String(5000);
  embedding : Vector(1536);
}

Use this with the HANA Cloud Vector Engine and AI Core orchestration grounding to build RAG scenarios directly in your CAP data model.

Prompt Externalization

Do not hardcode prompts in event handlers. Store them in JSON files or a CDS configuration entity so they can be updated without redeployment:

entity PromptTemplates {
  key id       : UUID;
  name         : String(100);
  systemPrompt : LargeString;
  updatedBy    : String;
  modifiedAt   : Timestamp;
}

Memory and Deployment

Node.js containers with AI SDK processing large text payloads require at least 512MB memory in the MTA descriptor. The AI SDK and JSON payload handling consume more memory than typical CAP services.

For complete SDK documentation, see sap-cloud-sdk-ai skill. For AI Core platform setup and orchestration configuration, see sap-ai-core skill.

Database Setup

Development (SQLite)

// package.json
{
  "cds": {
    "requires": {
      "db": {
        "[development]": { 
          "kind": "sqlite", 
          "credentials": { "url": ":memory:" } 
        },
        "[production]": { "kind": "hana" }
      }
    }
  }
}

Production (SAP HANA)

cds add hana
cds deploy --to hana

Initial Data (CSV)

  • File location: db/data/my.bookshop-Books.csv
  • Format: <namespace>-<EntityName>.csv
  • Auto-loaded on deployment

Deployment

Cloud Foundry

# Add CF deployment support
cds add hana,xsuaa,mta,approuter

# Build and deploy
npm install --package-lock-only
mbt build
cf deploy mta_archives/<project>_<version>.mtar

Multitenancy (SaaS)

cds add multitenancy

Configuration:

{
  "cds": {
    "requires": {
      "multitenancy": true
    }
  }
}

Authorization Examples

// Service-level
@requires: 'authenticated-user'
service CatalogService { ... }

// Entity-level
@restrict: [
  { grant: 'READ' },
  { grant: 'WRITE', to: 'admin' }
]
entity Books { ... }

Bundled Resources

Reference Documentation (22 files)

  1. references/annotations-reference.md - Complete UI annotations reference (10K lines)
  2. references/cdl-syntax.md - Complete CDL syntax reference (503 lines)
  3. references/cql-queries.md - CQL query language guide
  4. references/csn-cqn-cxn.md - Core Schema Notation and query APIs
  5. references/data-privacy-security.md - GDPR and security implementation
  6. references/databases.md - Database configuration and deployment
  7. references/deployment-cf.md - Cloud Foundry deployment details
  8. references/event-handlers-nodejs.md - Node.js event handler patterns
  9. references/extensibility-multitenancy.md - SaaS multitenancy implementation
  10. references/fiori-integration.md - Fiori Elements and UI integration
  11. references/java-runtime.md - Java runtime support
  12. references/localization-temporal.md - i18n and temporal data
  13. references/nodejs-runtime.md - Node.js runtime reference
  14. references/plugins-reference.md - CAP plugins and extensions
  15. references/tools-complete.md - Complete CLI tools reference
  16. references/consuming-services-deployment.md - Service consumption patterns
  17. references/service-definitions.md - Service definition patterns
  18. references/event-handlers-patterns.md - Event handling patterns
  19. references/cql-patterns.md - CQL usage patterns
  20. references/cli-complete.md - Complete CLI reference
  21. references/mcp-integration.md - MCP server setup and usage guide (new)
  22. references/mcp-use-cases.md - Illustrative MCP workflow scenarios (new)

Templates (8 files)

  1. templates/bookshop-schema.cds - Complete data model example
  2. templates/catalog-service.cds - Service definition template
  3. templates/fiori-annotations.cds - UI annotations example
  4. templates/mta.yaml - Multi-target application descriptor
  5. templates/package.json - Project configuration template
  6. templates/service-handler.js - Node.js handler template
  7. templates/service-handler.ts - TypeScript handler template
  8. templates/xs-security.json - XSUAA security configuration

Quick References

Common CLI Commands

cds init [name]           # Create project
cds add <feature>         # Add capability
cds watch                 # Dev server with live reload
cds serve                 # Start server
cds compile <model>       # Compile CDS to CSN/SQL/EDMX
cds deploy --to hana      # Deploy to HANA
cds build                 # Build for deployment
cds env                   # Show configuration
cds repl                  # Interactive REPL
cds version               # Show version info

Best Practices

DO ✓

  • Use cuid and managed aspects from @sap/cds/common
  • Keep domain models in db/, services in srv/, UI in app/
  • Use managed associations (let CAP handle foreign keys)
  • Design single-purpose services per use case
  • Start with SQLite, switch to HANA for production

DON'T ✗

  • Don't use SELECT * - be explicit about projections
  • Don't bypass CAP's query API with raw SQL
  • Don't create microservices prematurely
  • Don't hardcode credentials in config files
  • Don't write custom OData providers

Version Information

  • CAP Version: @sap/cds 9.7.x
  • MCP Version: @cap-js/mcp-server 0.0.5
  • LSP Version: @sap/cds-lsp 9.7.x
  • License: GPL-3.0