palarran-infrastructure-incident

原文涵盖的主题

• /palarran:infrastructure:incident

Read incident response, coordination, sources, security, operations and access workflow before decisions. Follow global memory and decision memory; an unavailable memory service must not delay authorized incident response. --plan stays offline and writes only inert local investigation artifacts. 1. Resolve the affected project/environment, business impact, time window and actual access/change scope. Read target AGENTS, incident runbook, delivery state and relevant decisions. Missing facts stay unknown. Do not require a new feature branch/worktree before diagnosis or an already-authorized emergency mitigation; permanent code fixes use the normal feature workflow separately. 2. Establish one incident ID and coordinator, with an operations owner and independent recovery verifier. Several developers/machines must join the same project/incident room using distinct identities. Discover the actual available coordination tools and capability/version; never invent a chat connection or cross-machine lock. If absent, use an explicitly agreed single operator and coordinator-owned journal; native subagent messaging alone does not synchronize other developer sessions. 3. Delegate independent, bo

这里只展示节选，请以完整原文和当前要求为准。