{"version":"2.1.287","anchor":"remote-background-shell-tool-results-tell-the-model-how-to-c","canonical_anchor":"remote-background-shell-tool-results-tell-the-model-how-to-c","heading":"Claude is told how to check and stop remote background commands safely","tier":"notice","area":"Bash Tool","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287\/e\/remote-background-shell-tool-results-tell-the-model-how-to-c","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287","markdown":"### Claude is told how to check and stop remote background commands safely\n\nFor remote background commands with a task id, Claude is now told which tools to use to check and stop them, and not to use pgrep, pkill or kill\n\n**Unclear.** It is not clear what condition decides when the new message is used.\n\n**What**\n\nWhen Claude starts a shell command that keeps running in the background, Claude Code sends Claude a short message about it. When that background task has a task id, the message now tells Claude to check on the task with one tool and stop it with another, and warns it never to check on it or stop it with `pgrep`, `pkill` or `kill`. When there is no task id, Claude gets the same generic message as before, which only says the task id and output file exist.\n\nThis new wording is behind a check that decides when it is used.\n\n**Why**\n\nThe change alters what Claude is told about its own background commands, steering it toward the tools built for the job rather than ordinary process commands.\n\n- Area: Bash Tool\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 3\/5"}