{"version":"2.1.296","anchor":"the-agent-tool-description-was-rewritten","canonical_anchor":"the-agent-tool-description-was-rewritten","heading":"Agent tool instructions now call the brief Claude's first lever, not its only one","tier":"internal","area":"Subagents","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.296\/e\/the-agent-tool-description-was-rewritten","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.296","markdown":"### Agent tool instructions now call the brief Claude's first lever, not its only one\n\nClaude's instructions for handing work to a subagent now call the brief its \"first\" lever on cost instead of its \"one\" lever\n\n**What**\n\nThe Agent tool lets Claude start a subagent, a helper that works on part of a task with its own separate context. The description that tells Claude how to use it differs in the v2.1.296 capture from the v2.1.295 one, for several of the models Claude Code can use.\n\nThe guidance on briefing a subagent keeps its advice: state the goal and what has been ruled out, point to the files worth reading, and keep the scope narrow. The brief was called Claude's \"one lever\" on the costs of a subagent. It is now called Claude's \"first lever\".\n\n**Why**\n\nThese are Claude's own instructions, so their wording shapes how it hands work off. The new wording still stresses a clear brief but no longer presents it as the only thing Claude can adjust.\n\n- Area: Subagents\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 0\/5\n- Scope: individual\n- Heads-up: no"}