Detects the first-time installation of a Codex plugin for a specific plugin_display_name/plugin_name pair in the OpenAI ChatGPT Enterprise Codex logs, within the rule history window. This rule flags newly added Codex plugins that can introduce connectors to access mail, chat, or files inside Codex sessions, which adversaries might abuse for persistence or data access after initial compromise. It triggers on plugin_installed events in the openai_chatgpt_enterprise.codex_logDataset, requiring plugin_display_name and plugin_name fields. The rule includes investigation steps, false positives (approved rollout of common plugins like Slack, Gmail, Drive), triage guidance, and remediation actions. It maps to MITRE ATT&CK techniques T1176 (Software Extensions) for Persistence and T1213 (Data from Information Repositories) for Collection, and to MITRE ATLAS AML.T0053 (AI Agent Tool Invocation) and AML.T0085 (Data from AI Services). Setup requires enabling Codex compliance logs in Elasticsearch. Severity is medium, with a risk_score of 47. Investigative fields include organization.id, user.id, user.email, and plugin details to support attribution and validation of rollout vs. anomaly.