Prompt-Cache-Aware Conversation Design

Structure your agent's context for prefix caching: stable prefixes, append-only history, and cache-friendly dynamic content.

0 runs0 forks0

Chain

Inputs

Step 1

flow-prompt-cache-aware-conversation-design-1-audit-context-stability

Step 2

flow-prompt-cache-aware-conversation-design-2-restructure-for-stable-first-layou

Step 3

flow-prompt-cache-aware-conversation-design-3-verify-cache-hits-and-measure-savi

Final output

No inputs required — hit Run to stream each step in order.

Each step streams from its model and feeds the next. BYOK — you pay your provider directly. Add a key in Settings if a step needs one.

Step 1 · Find every byte that changes between turns - each one is a cache killer.

flow-prompt-cache-aware-conversation-design-1-audit-context-stability

Claude Sonnet 4.6→ {{step_1}}

Awaiting upstream steps…

Step 2 · The layout rule: immutable, then append-only, then volatile.

flow-prompt-cache-aware-conversation-design-2-restructure-for-stable-first-layou

Claude Sonnet 4.6→ {{step_2}}

Awaiting upstream steps…

Step 3 · The API tells you the truth - read it.

flow-prompt-cache-aware-conversation-design-3-verify-cache-hits-and-measure-savi

Claude Sonnet 4.6→ {{step_3}}

Awaiting upstream steps…

Structure your agent's context for prefix caching: stable prefixes, append-only history, and cache-friendly dynamic content.

1
Find every byte that changes between turns - each one is a cache killer.
Runs flow-prompt-cache-aware-conversation-design-1-audit-context-stability on Claude Sonnet 4.6
2
The layout rule: immutable, then append-only, then volatile.
Runs flow-prompt-cache-aware-conversation-design-2-restructure-for-stable-first-layou on Claude Sonnet 4.6
3
The API tells you the truth - read it.
Runs flow-prompt-cache-aware-conversation-design-3-verify-cache-hits-and-measure-savi on Claude Sonnet 4.6

•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.

Claude Sonnet 4.6