Skip to content
sinc-LLMBook a 30-minute call

Claude Prompt Best Practices: The 6-Band Framework

Structured prompt fields and Claude-specific examples, with a discussion of prompt design and API costs.

On this page
By Mario Alexandre March 21, 2026 sinc-LLM Prompt Engineering

Anthropic's Claude models are good at following instructions, staying safe, and reading long text. These traits make Claude easy to guide. When you give Claude clear rules, it follows them closely.

This guide discusses the sinc-LLM six-field format with Claude-based prompting examples. Related: sinc-LLM framework.

The 6 Bands for Claude

x(t) = Σ x(nT) · sinc((t - nT) / T)

Here are the 6 specification bands, with Claude-specific notes from my testing:

  1. PERSONA, Claude does well when you give it a clear role. "You are a distributed systems architect" gets better results than "You are helpful."
  2. CONTEXT, Claude can read up to 200K tokens at once. That is a huge window. But more is not always better. Band decomposition helps you include only the context that matters.
  3. DATA, Claude handles structured data well. Give your data in clean formats like JSON, CSV, or markdown tables.
  4. CONSTRAINTS, This is where Claude shines. Claude follows detailed rules precisely. Put 42% of your prompt tokens here.
  5. FORMAT, Claude gives neat, consistent output when you show it exactly what format you want. Use a sample of the output you expect.
  6. TASK, Keep it short. Claude does not need a long task description when the other 5 bands are set up well.

Use XML Tags for Band Separation

Claude works especially well with XML tags around each section. Wrap each band in clear tags like this:

<persona>Senior security engineer</persona>
<context>Production Kubernetes cluster, 50 services</context>
<data>[pod logs from the last 2 hours]</data>
<constraints>
- Focus on network policy violations only
- Do not suggest changes to application code
- Flag any pod-to-pod communication not in the allowlist
</constraints>
<format>Table: Pod | Violation | Severity | Recommendation</format>
<task>Audit these logs for network policy violations.</task>

Leverage Claude's Thinking

For complex tasks, add this rule: "Think step by step before answering, but show only the final output." This turns on Claude's extended thinking mode without adding clutter to the result.

Cost Optimization for Claude API

Claude API costs depend on the selected model, token usage, and other billing conditions. Compare prompt changes against the same workload and output checks:

MetricUnstructured Prompt6-Band Prompt
Input tokens80,0002,500
Output quality (SNR)0.0030.92
Retry rate~30%~2%
Effective cost per task$0.52$0.016

Resources

These are the tools and references I suggest for Claude prompt work:

Review a prompt using six specification fields

Try sinc-LLM Free

This is the exact JSON format that sinc-LLM uses. Paste any raw prompt at sincllm.com to generate one automatically.

{
  "formula": "x(t) = Σ x(nT) · sinc((t - nT) / T)",
  "T": "specification-axis",
  "fragments": [
    {
      "n": 0,
      "t": "PERSONA",
      "x": "You are a Claude API specialist with 2 years of production experience. You provide precise, evidence-based analysis with exact numbers and no hedging."
    },
    {
      "n": 1,
      "t": "CONTEXT",
      "x": "This analysis is part of a production system where accuracy determines revenue. The sinc-LLM framework identifies 6 specification bands with six named fields."
    },
    {
      "n": 2,
      "t": "DATA",
      "x": "Example allocation assumptions: CONSTRAINTS=42.7%, FORMAT=26.3%, PERSONA=7.0%, CONTEXT=6.3%, DATA=3.8%, TASK=2.8%. SNR formula: 0.588 + 0.267 * G(Z1) * H(Z2) * R(Z3) * G(Z4). These allocations are illustrative assumptions, not observed quality weights."
    },
    {
      "n": 3,
      "t": "CONSTRAINTS",
      "x": "State facts directly. Never hedge with 'I think' or 'probably'. Use exact numbers for every claim. Do not suggest generic solutions. Every recommendation must be specific and verifiable. Include at least 3 MUST/NEVER rules specific to this task."
    },
    {
      "n": 4,
      "t": "FORMAT",
      "x": "Lead with the definitive answer. Use structured headers. Tables for comparisons. Numbered lists for sequences. Code blocks for implementations. No trailing summaries."
    },
    {
      "n": 5,
      "t": "TASK",
      "x": "Optimize a Claude system prompt for a customer support bot using all 6 sinc bands"
    }
  ]
}

Install: pip install sinc-llm | GitHub | Paper

// Production AI Engineering

Build AI systems that hold up in production.

sinc-LLM designs, audits, and stabilises production AI infrastructure: from vendor evaluation and cost accountability to incident controls and MCP architecture.

See what we do →
Additional context

sinc-LLM About Blog Services

The services, methods, and examples described here do not establish an expected performance result for your workload.