{"version":"2.1.283","anchor":"prompt-hook-evaluator-agents-tools-are-filtered-by-permissi","canonical_anchor":"prompt-hook-evaluator-agents-tools-are-filtered-by-permissi","heading":"Hook-checking agents can only use tools your permission mode allows","tier":"notice","area":"Hooks","url":"https:\/\/changelogs.core-directive.com\/v\/2.1.283\/e\/prompt-hook-evaluator-agents-tools-are-filtered-by-permissi","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.283","markdown":"### Hook-checking agents can only use tools your permission mode allows\n\nThe agent that checks prompt and agent hooks now loses any tool the session's current permission mode does not allow\n\n**Unclear.** The exact check used to decide which tools the permission mode disallows is not described.\n\n**What**\n\nSome hooks (checks that run at set points, such as when Claude is about to stop) are evaluated by a separate helper agent that reads the conversation and decides whether a condition is met. That helper already had its tools filtered. It now also drops any tool that the session's current permission mode does not allow. A permission mode is the setting that decides how freely Claude Code may act without asking you.\n\n**Why**\n\nA hook evaluator no longer gets tools that the session's own mode forbids.\n\n- Area: Hooks\n- Tier: You'll notice\n- Useful: 1\/5\n- Signal: 1\/5"}