{"version":"2.1.98","anchor":"agent-mode-command-validation-gradual-rollout","canonical_anchor":"agent-mode-command-validation-gradual-rollout","heading":"Agent Mode Command Validation [Gradual Rollout]","tier":null,"area":null,"scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.98\/e\/agent-mode-command-validation-gradual-rollout","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.98","markdown":"### Agent Mode Command Validation [Gradual Rollout]\n\n**What**\n\nValidation layer that blocks certain long-running synchronous bash commands when running in agent mode.\n\n**Status**\n\nFeature-flagged via `tengu_amber_sentinel`\n\n**Details**\n\n- When enabled, checks commands before execution in foreground mode\n- Blocks standalone `sleep` commands with duration \u2265 2 seconds\n- Designed to prevent agents from wasting time on unnecessary blocking operations\n- Commands can bypass the check by using `run_in_background: true`\n\n**Evidence**\n\nAgent command validation (gated by `tengu_amber_sentinel`, search for `\"standalone sleep\"`)\n\n- Flag `tengu_amber_sentinel`: On for this account, and not off by default (read for one account on one subscription tier against v2.1.98; this account: on, anonymous baseline: on, compiled default: no gate table built for this version) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release."}