Forge

Your infrastructure runs on tribal knowledge. You are the only person who remembers why that security group rule exists, why that service mesh config has a hardcoded timeout, and why production deploys happen on Tuesday nights because "traffic is low." Every incident pages you at 2am because your runbooks are three months stale and your on-call engineer cannot find the rollback procedure. You have tried wikis, Notion, and Confluence. They are out of date the moment the incident starts. Forge is a DevOps and infrastructure intelligence architect built to run the operational layer of engineering organizations that have outgrown manual processes. Extracted from 200+ real incident response cycles, 40+ production Terraform codebases, and live Kubernetes fleet management across multi-cloud environments. Every pattern in Forge's behavior exists because a real production system failed in a predictable way. The anti-patterns Forge refuses to execute are the exact mistakes that caused those failures. Unlike a generic engineering assistant, Forge never applies infrastructure changes directly to production without a validated plan file and an explicit rollback procedure documented first — because "it worked in staging" has taken down production hundreds of times, and the cost is never just the downtime. That rule exists because Terraform state drift during an active incident is unrecoverable without version-pinned state backups. Forge enforces a plan-first, rollback-documented, blast-radius-estimated workflow on every infrastructure change, regardless of urgency or stakeholder pressure. What you get: SOUL.md — Forge's operational identity, decision-making framework, and escalation logic. IDENTITY.md — role definition, expertise boundaries, and communication protocols for engineering teams. INFRA_OPS.md — Terraform, Kubernetes, and cloud provider workflow playbooks with environment-specific decision trees. INCIDENT_RESPONSE.md — structured runbook templates, severity classification matrix, and postmortem framework. RELEASE_PIPELINE.md — zero-downtime deployment patterns, canary release logic, and rollback trigger conditions. Requires cloud provider CLI access (AWS, GCP, or Azure), Terraform, and a Kubernetes toolchain.