What problem does it solve? Taking an AI agent from a local prototype to production requires choosing among many hosting stacks and executing dozens of error-prone steps involving cloud consoles, secrets, and billing. This Skill removes the guesswork by profiling the agent's actual capabilities, deterministically scoring seven deployment stack archetypes, and walking the user through a guided go-live runbook. ## Core Features & Use Cases - Capability Detection: Delegates to a repo-capability-scanner subagent to detect UI type, run duration, state requirements, auth mode, and other deployment-relevant capabilities, then confirms findings with the user in one pass. - Deterministic Stack Scoring: Runs scripts/score_stack.py, a direct port of the decision guide's scoring engine, to rank seven stacks (Streamlit hosting, PaaS, Cloud Run, Vercel, VM, managed hyperscaler, Kubernetes) as recommended, viable, or poor fit. - Guided Runbooks: Executes stack-specific runbooks from references/ with [AUTO] steps performed by the agent and [HUMAN] steps that pause for console actions, credentials, or spending decisions. - Use Case: A developer finishes a LangGraph agent with local SQLite state and asks how to ship it. The skill detects the local state and streaming needs, recommends Cloud Run or a VM, then guides provisioning, secrets, Docker, and rollback step by step. ## Quick Start Invoke /deploy-agent with the path to your agent project to get a stack recommendation and a guided deployment walkthrough.
