Detects OpenAI ChatGPT Enterprise custom-agent publish events (agent_published) emitted by the OpenAI integration. Publishing a custom agent that includes connectors, skills, or store visibility can create durable automation and broaden data access beyond the publishing user. The rule ingests compliance logs from the OpenAI ChatGPT Enterprise integration (logs-openai_chatgpt_enterprise.custom_agents_log*) and triggers on event.action == agent_published, surfacing fields such as user.id, user.email, gen_ai.agent.id, publish_type, release_name, agent_fields.store_visible, and agent_fields.workspace_capability. It relies on the integration to forward logs into Elasticsearch. MITRE ATT&CK mapping includes T1176 Software Extensions under Persistence, and MITRE ATLAS AML.T0081 Modify AI Agent Configuration under Persistence. Triage guides escalate when publish scope or connectors are broader than expected or the publisher is not authorized; false positives occur for approved platform rollouts. Remediation involves unpublishing or deleting unauthorized agents and revoking related OAuth connections, then reviewing recent agent runs if applicable. The setup requires enabling the OpenAI ChatGPT Enterprise integration to collect compliance logs into Elasticsearch. References include Elastic OpenAI ChatGPT Enterprise docs and OpenAI compliance guidance.