Claudelicious

Claudelicious is the wiring diagram behind one operator's full Claude Code setup, published as an open cookbook.

The idea: Most "here is my setup" repos are a flat list of skills. Claudelicious leads with the systems underneath the skills, because those are the part you cannot reverse-engineer from a screenshot: a harness that remembers across sessions, improves itself when you correct it, and runs while you sleep, all wired around a homelab.

What it covers (each doc: principle, then a worked example, then ship-or-scrub notes)

• Rules and context that stop you repeating instructions every session
• A skill library that stays small and sharp
• Hooks that stop the agent doing dangerous things
• Memory + the learning loop, so a correction sticks instead of recurring
• Continuity across compactions
• Session search (mneme) and a semantic second brain over your notes
• Multi-provider routing (Kimi / MiniMax / GLM / Qwen) and scheduled jobs
• Always-on agents that run on a heartbeat
• Voice and anti-slop, visual generation, subagents and workflows, the fleet, MCP

The five words it builds on: Model, Harness, Agent, Skill, MCP. A great model with no harness is a demo. A harness with a great model is a second nervous system.