CVE-2026-65699 is an authorization bypass vulnerability in reworkd's AgentGPT product through version 1.0.0. The issue occurs because the application accepts a user-supplied run_id in the request body to attach tasks to an agent run but fails to verify that the run_id belongs to the authenticated user making the request. This lack of ownership verification in the AgentCRUD.create_task and validate_task_count functions enables an attacker who obtains a valid run_id to interfere with another user's agent run by corrupting task history, exhausting the run's loop budget, and driving up costs associated with the large language model (LLM) usage on the victim's account.

An attacker with a valid run_id can manipulate another user's agent run by attaching unauthorized tasks. This can lead to corrupted task histories, depletion of the allowed loop budget per run, and increased operational costs for the victim due to excessive LLM usage. The CVSS 4.0 base score is 2.3, indicating a low severity impact primarily due to the requirement for authenticated access and high attack complexity.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is available, restrict access to valid run_ids and monitor for unusual task attachment activity. Users should avoid sharing run_ids and ensure that only authorized users have authenticated access. Follow vendor updates for official patches or fixes.