{"version":"2.1.302","anchor":"tool-prompt-cache-key-gains-a-bit-for-agent-whentouse","canonical_anchor":"tool-prompt-cache-key-gains-a-bit-for-agent-whentouse","heading":"Cached tool descriptions now depend on whether an agent has a 'when to use' text","tier":"internal","area":"Elsewhere","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.302\/e\/tool-prompt-cache-key-gains-a-bit-for-agent-whentouse","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.302","markdown":"### Cached tool descriptions now depend on whether an agent has a 'when to use' text\n\nThe key Claude Code uses to reuse cached tool text now changes depending on whether an agent's 'when to use' description is empty\n\n**What**\n\nClaude Code saves time by reusing text it has already sent to the model, which is called prompt caching. The key that decides whether a tool's text can be reused now includes one more piece of information: whether an agent's 'when to use' description is empty or filled in.\n\n**Why**\n\nIn practice, cached tool text is now kept separate for agents with and without that description.\n\n- Area: Elsewhere\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}