Anth Reference Architecture

What it does

When your agent uses it

• Designing a Claude-powered application
• Choosing between direct API vs queue-based
• Planning a multi-model architecture
• With phrases like anthropic architecture

Example prompts

• “anthropic architecture”
• “claude system design”
• “anthropic reference architecture”
• “/anth-reference-architecture”

Requirements

• Python 3
• Compatibility (from SKILL.md): Designed for Claude Code
• Pre-approved tools (allowed-tools): Read, Write, Edit, Grep

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

1. 1Select the smallest architecture that meets the workload: gateway for interactive calls, queue for asynchronous work, or a router only…
2. 2Keep credentials and policy enforcement at the service boundary. Validate model, token, rate, data-class, source, and destination scope…
3. 3Exercise success, timeout, 429/5xx, duplicate, queue-retry, tool-use, and partial-response paths with synthetic fixtures. Ensure traces…
4. 4Canary the selected topology in an isolated workspace, observe SLOs/cost/rate limits, and require approval before production traffic…
5. 5On policy, reliability, or cost regression, open the circuit or pause workers, drain/quarantine unsafe work, roll back, and retain a…

What it can do on your machine

Read from SKILL.md and the folder at commit cfae287. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Read
  • Write
  • Edit
  • Grep

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are python).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  Links to these hosts (documentation or services it may open):

  • platform.claude.com

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

• Compatibility

  Designed for Claude Code

  From compatibility in the SKILL.md frontmatter.

Context cost

Anth Reference Architecture loads about 1.8k tokens when it runs. Until then it costs about 92 tokens; SKILL.md has 327 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~92

When it runs · the whole SKILL.md, loaded when a task matches

~1.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

• NoteMentions a .env fileSKILL.md:153

  ├── .env.development

• NoteMentions a .env fileSKILL.md:154

  ├── .env.staging

• NoteMentions a .env fileSKILL.md:155

  └── .env.production

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from jeremylongshore/tons-of-skills-marketplace at commit cfae287, republished under its MIT licence (© jeremylongshore). 327 words, ~1,798 tokens.

name

anth-reference-architecture

description

Implement Claude API reference architectures for common use cases. Use when designing a Claude-powered application, choosing between direct API vs queue-based, or planning a multi-model architecture. Trigger with phrases like "anthropic architecture", "claude system design", "anthropic reference architecture", "design claude integration".

allowed-tools

Read, Write, Edit, Grep

compatibility

Designed for Claude Code

version

1.7.0

license

MIT

author

Jeremy Longshore <jeremy@intentsolutions.io>

tags

saas, ai, anthropic

User → API Gateway → Claude Service → Messages API
                                     ↓
                                   Response → User

# Best for: chatbots, interactive tools, low-volume (<100 RPM)
from fastapi import FastAPI
import anthropic

app = FastAPI()
client = anthropic.Anthropic(max_retries=3, timeout=60.0)

@app.post("/chat")
async def chat(prompt: str):
    msg = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=1024,
        messages=[{"role": "user", "content": prompt}]
    )
    return {"text": msg.content[0].text, "tokens": msg.usage.output_tokens}

User → API → Queue (Redis/SQS) → Worker Pool → Messages API
  ↑                                                ↓
  └──────────── Status/Result ←── Result Store ←───┘

# Best for: batch processing, high-volume, background tasks
from redis import Redis
from rq import Queue
import anthropic

redis = Redis()
task_queue = Queue("claude-tasks", connection=redis)
result_store = Redis(db=1)

def process_task(task_id: str, prompt: str, model: str):
    client = anthropic.Anthropic()
    msg = client.messages.create(
        model=model,
        max_tokens=1024,
        messages=[{"role": "user", "content": prompt}]
    )
    result_store.setex(f"result:{task_id}", 3600, msg.content[0].text)

# Enqueue
import uuid
task_id = str(uuid.uuid4())
task_queue.enqueue(process_task, task_id, prompt, "claude-sonnet-4-20250514")

User → Router → Haiku    (classify/extract)
              → Sonnet   (general/code)
              → Opus     (research/complex)
              → Batches  (bulk/offline)

class ModelRouter:
    def __init__(self):
        self.client = anthropic.Anthropic()
        self.classifier = anthropic.Anthropic()  # Can be same client

    def route_and_execute(self, prompt: str, context: dict) -> str:
        # Step 1: Classify with Haiku (cheap, fast)
        classification = self.classifier.messages.create(
            model="claude-haiku-4-20250514",
            max_tokens=32,
            messages=[{
                "role": "user",
                "content": f"Classify this request as: simple|moderate|complex|bulk\n\n{prompt[:200]}"
            }]
        )
        complexity = classification.content[0].text.strip().lower()

        # Step 2: Route to appropriate model
        model_map = {
            "simple": "claude-haiku-4-20250514",
            "moderate": "claude-sonnet-4-20250514",
            "complex": "claude-opus-4-20250514",
        }
        model = model_map.get(complexity, "claude-sonnet-4-20250514")

        # Step 3: Execute with selected model
        msg = self.client.messages.create(
            model=model,
            max_tokens=4096,
            messages=[{"role": "user", "content": prompt}]
        )
        return msg.content[0].text

my-claude-app/
├── src/
│   ├── main.py              # FastAPI app
│   ├── claude/
│   │   ├── client.py         # Singleton + config
│   │   ├── router.py         # Model routing logic
│   │   ├── tools.py          # Tool definitions
│   │   └── prompts/          # System prompts as files
│   ├── workers/
│   │   └── claude_worker.py  # Queue consumer
│   └── middleware/
│       ├── rate_limiter.py   # App-level rate limiting
│       └── cost_tracker.py   # Spend monitoring
├── tests/
│   ├── unit/                 # Mocked tests
│   └── integration/          # Live API tests
└── config/
    ├── .env.development
    ├── .env.staging
    └── .env.production

© jeremylongshore, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/.curated/anth-reference-architecture of jeremylongshore/tons-of-skills-marketplace.

Open the folder on GitHubat commit cfae287