AI Engineering

Journey Agents builds the agents listed in 02-plan.md by delegating each one to the agents skill, then registers any required MCP servers, validates the specs, creates the agents, and smokes them. Use it during the agents stage of a Butterbase journey.

Journey AI configures the AI stage of the Butterbase journey by setting gateway defaults and optional BYOK. Use it when a plan includes LLM or embeddings setup and the AI section needs to be applied and smoke-tested.

Journey Auth configures OAuth providers and optional auth hooks as the auth stage of a Butterbase build journey. Use it after schema setup when an app needs end-user sign-in.

Journey Idea turns a rough app concept into a structured brief by asking one concrete question at a time. It captures the first user, first action, must-haves, and Butterbase capability tags in docs/butterbase/01-idea.md.

Journey Plan turns a Butterbase idea brief and capability map into a concrete app plan. Use it to define tables, auth, functions, storage, AI features, frontend stack, and hackathon scope cuts.

Journey RAG creates RAG collections and ingests seed documents as stage 3g of the Butterbase guided journey. Use it when the plan includes a knowledge base and you need to prepare retrieval data and verify it with a smoke query.

WeChat Article Pipeline orchestrates a WeChat public account content flow from topic selection through writing, review, formatting, images, and publishing. Use it for one-step creation or to continue an existing article workflow.

Maestro Knowhow captures reusable knowhow into .workflow/knowhow/ from an intent you provide. It infers the content type, extracts the content, and writes the entry using the workflow instructions.

Maestro Knowledge manages an intent-driven knowledge store for auditing, reviewing, staging, promoting, and harvesting knowledge. Use it when you want Maestro to route a knowledge lifecycle task from a natural-language intent or explicit keyword.

Skill Generator creates new Claude Code skills with a configurable sequential or autonomous execution model. Use it when scaffolding a skill, defining its phases, or building a new workflow.

Skill Iter Tune runs a skill through execute-evaluate-improve loops, using Claude to execute, Agy to evaluate, and Agent to apply changes. Use it when you want to refine a skill until it reaches a quality threshold or hits a max iteration limit.

Skill Tuning diagnoses and repairs Claude Code skill execution problems such as context explosion, memory loss, data flow disruption, and coordination failures. Use it when a skill is producing inconsistent results or needs optimization.