Проектировщик агентных воркфлоу GitHub Actions

★ 6.5 · ai-agents

agentic-workflow-designer is a Claude Code skill that runs a structured conversational interview with users and generates a complete GitHub Agentic Workflow `.md` file based on their answers. The interview proceeds through eight sequential phases — automation goal, trigger, read/write scope, data strategy, guardrails, external APIs and MCP servers, AI engine preference, and a confirmation summary — advancing only when each phase is fully resolved. Built-in mapping tables translate natural-language descriptions into workflow syntax: `on:` triggers, `permissions:`, `tools:`, `safe-outputs:`, `network.allowed`, and secrets wiring for integrations such as Slack, Jira, or custom MCP servers, with support for GHES and GHEC environments. It is designed for developers who know what they want to automate but are unfamiliar with the workflow syntax, and should be used before `create-agentic-workflow.md` when requirements are still unclear.