Assemble a System Prompt Like the Pros

About this workflow

Structure your agent's system prompt the way leading harnesses do: layered sections, dynamic context, and enforceable rules.

What happens when you run it

1. 1
   Name every instruction's job before deciding where it lives.

   Runs flow-assemble-system-prompt-pros-1-decompose-the-current-prompt on Claude Sonnet 4.6
2. 2
   Prompts as code: composable, reviewable, and measurable.

   Runs flow-assemble-system-prompt-pros-2-build-the-section-based-assembler on Claude Sonnet 4.6
3. 3
   A rule the harness cannot check is a wish.

   Runs flow-assemble-system-prompt-pros-3-pair-rules-with-enforcement on Claude Sonnet 4.6
4. 4
   Measure the prompt like a system, not a spell.

   Runs flow-assemble-system-prompt-pros-4-ablate-to-prove-section-value on Claude Sonnet 4.6

Best for

• •Recurring tasks you currently copy-paste between tools.
• •When a single prompt isn't enough — you need a chain where output(n) feeds prompt(n+1).
• •Fork and adapt if your inputs differ materially from the author's use case.

Models used