Long-form runbooks for the patterns I've shipped. The why, the how, and the gotchas — written for humans, with screenshots when they help. Each recipe pairs with an installable Agent Skill: the recipe explains it; the skill makes it executable.
Companion essay: Skills as SOPs.
An Obsidian vault at ~/brain wired to Claude Code as persistent memory. SessionStart and Stop hooks, a /brain skill that saves notes from any session, an auto-committing git repo, and a semantic search index. The pattern that keeps me from forgetting what I learned the last time.
A synchronous, speaking avatar embedded in a webpage using HeyGen and ElevenLabs. Cloudflare Worker token exchange, SDK integration, mobile gotchas, and when not to use it.
Read the recipe →Connect external documents to your ElevenLabs Conversational AI agent so it retrieves relevant context before answering. Knowledge-base API calls, vectorization, and the GitHub page_build gotcha.
Let an ElevenLabs voice agent invoke functions that run inside the visitor's browser — open modals, scroll to sections, play videos — by wiring client-side tool calls to the page's JavaScript.
Read the recipe →Centralize every scattered API key into one protected user-level store at ~/.config/dev-secrets/. Audit existing .env files, migrate by layer, revoke leaked tokens, and never paste a key in chat again.
Recipes pulled from the brain vault as patterns prove themselves on real projects. Tracked queue: