{"version":"2.1.232","anchor":"recovered-web-fetch-agents-are-resumed-by-message-not-by-fil","canonical_anchor":"recovered-web-fetch-agents-are-resumed-by-message-not-by-fil","heading":"Recovered web-fetch agents are resumed by message, not by file","tier":"soon","area":"Web Fetch","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.232\/e\/recovered-web-fetch-agents-are-resumed-by-message-not-by-fil","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.232","markdown":"### Recovered web-fetch agents are resumed by message, not by file\n\nRecovered fetch agents are reached by message rather than by reading an output file.\n\n**What**\n\nBackground agents left running by a previous Claude Code process are now marked as web-fetch launches. Those agents have no worktree and no output file, so the recovery message tells the model to reach them with SendMessage rather than to read files, including asking it to message a finished agent to get its report. Only populated when the web-fetch agent is enabled.\n\n**Details**\n\n- The multi-agent summary uses \"fetched web content and has\" or the plural form depending on how many were recovered.\n\n**Evidence**\n\n`fetched web content and has`\n\n- Area: Web Fetch\n- Tier: Nothing to try yet\n- Useful: 2\/5\n- Signal: 3\/5\n- Present in the build but not switched on"}