Agent skill

ucp-dev-patterns

UCP development patterns — capability negotiation algorithms, idempotency implementation, error resolution loops, multi-binding servers, and production architecture. Use when designing the internal architecture of a UCP implementation or solving cross-cutting concerns.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/ucp-dev-patterns

SKILL.md

UCP Development Patterns

Before writing code

Fetch live reference: Web-search site:ucp.dev specification reference for the latest data model definitions and enum values. Also check https://ucp.dev/2026-01-23/documentation/core-concepts/ for architectural guidance.

Pattern: Capability Negotiation

Every request/response pair involves negotiation:

  1. Platform sends UCP-Agent header with its profile URI
  2. Business fetches the platform profile (cache it — don't fetch on every request)
  3. Business computes the intersection: capabilities both support
  4. Orphaned extensions are pruned (extensions whose parent capability is not in the intersection)
  5. Business returns the negotiated ucp object

Caching: Cache the platform profile with a TTL (e.g., 5 minutes). Don't re-fetch on every request.

Version compatibility: If platform version > business version, return version_unsupported. If platform version <= business version, process using the platform's version semantics.

Pattern: Idempotency

Every mutating operation uses Idempotency-Key (UUID):

  1. Hash the key and check your cache/store
  2. If found: return the cached response (same status code, same body)
  3. If not found: process the request, cache the response keyed by the idempotency key
  4. Cache duration: 24+ hours minimum

Storage: Use Redis, database table, or in-memory store depending on scale. Key = idempotency_key, Value = serialized response + status code.

Race condition: If two identical requests arrive simultaneously, use locking (database row lock, Redis SETNX) to ensure only one processes.

Pattern: Agent Error Resolution Loop

The core Platform pattern for driving checkout to completion:

1. Create checkout
2. LOOP:
   a. Inspect status
   b. If "incomplete": read messages, resolve recoverable errors, update checkout, goto 2
   c. If "requires_escalation": surface continue_url to user, wait, goto 2
   d. If "ready_for_complete": acquire payment credential, call complete
   e. If "complete_in_progress": payment is being processed, poll/wait, goto 2
   f. If "completed": done — extract order
   g. If "canceled": done — report to user
3. Timeout after N iterations or session expiry

Error resolution heuristics:

  • missing_shipping_address → prompt user for address, update checkout
  • missing_email → prompt user for email, update checkout
  • invalid_cart_items → item unavailable, suggest alternatives or remove
  • discount_code_invalid → inform user, remove code

Pattern: Multi-Binding Server

A single Business server can support REST + MCP + A2A simultaneously:

app/
├── core/
│   ├── checkout_service.py    # Shared business logic (binding-agnostic)
│   ├── negotiation.py         # Capability negotiation
│   └── models.py              # UCP data models (from SDK)
├── bindings/
│   ├── rest/
│   │   └── routes.py          # REST endpoints
│   ├── mcp/
│   │   └── tools.py           # MCP tool handlers
│   └── a2a/
│       └── agent.py           # A2A agent message handler
└── discovery.py               # /.well-known/ucp serving all bindings

The key: all bindings share the same core business logic. The binding layer only handles transport-specific concerns (HTTP headers vs JSON-RPC envelope vs A2A message parts).

Pattern: Monetary Calculations

  • All amounts are integers in minor currency units (cents). $29.99 = 2999.
  • Never use floating point for money.
  • Totals include typed entries: items_discount, subtotal, discount, fulfillment, tax, fee, total.
  • Recalculate all totals on every update (don't trust client-submitted totals).

Pattern: Webhook Reliability

Business (sender):

  • Implement exponential backoff: 1s, 2s, 4s, 8s, 16s... up to a max
  • Set a max retry count (e.g., 10 attempts over 24 hours)
  • Log failed deliveries for manual retry
  • Always send the full order entity, not deltas

Platform (receiver):

  • Respond 2xx immediately, process async
  • Deduplicate using event_id
  • Verify Request-Signature before processing
  • Handle out-of-order deliveries gracefully (use created_time to determine freshness)

Pattern: Testing Strategy

  1. Unit tests: Test negotiation logic, total calculations, error message generation
  2. Integration tests: Test each binding independently against your core logic
  3. Conformance tests: Run the official suite from https://github.com/Universal-Commerce-Protocol/conformance
  4. Playground validation: Use https://ucp.dev/playground/ for interactive flow testing
  5. End-to-end: Test full flows including payment handler integration

Ecosystem References

Always check these for the latest developments before major implementation decisions:

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results