Resources
A small, honest shelf: free things that are already usable, and paid systems when the delivery path exists.
If it is not packaged, tested, and reachable, it does not get a big shiny card here. Trust-first, not junk drawer.
The AI house manager I built so the household stops living in one person's head. It catches the open loops — school emails, receipts, appointments, meals, travel — prepares the decisions, drafts the boring admin, and stops before anything external without your yes.
Buy once. 12 months of updates at the founding price as the system improves.
Open source
Everything here is public, free, and running in my own setup before it got a repo. Most are Claude Code skills — install one with a single command, or read why I built it first.
Editorial diagrams for Claude Code. Reads your site's colors and fonts, then draws twenty-seven diagram types as self-contained HTML and inline SVG. No rounded-box Mermaid defaults.
npx skills add cathrynlavery/diagram-design
The ops layer that keeps long-running agents alive: health checks, repair scripts, watchdogs, update triage, and security scans.
npx skills add cathrynlavery/openclaw-ops
Gives Claude Code a second opinion from OpenAI Codex, with plan review wired in through hooks so it happens automatically instead of when you remember.
npx skills add cathrynlavery/codex-skill
Orchestrator plans, Codex writes the code. One task per commit, a test gate before each one, and exactly one PR at the end.
npx skills add cathrynlavery/codex-build
A local session-mining loop that reads your own transcripts and stages agent improvements and content ideas from what actually happened.
npx skills add cathrynlavery/agent-improvement-loop
Turns hundreds of untitled Apple Voice Memos into a searchable archive with transcripts, summaries, and real titles. Runs on your machine, no API key.
npx skills add cathrynlavery/voice-memo-organizer
Builds a persistent map of any git repo — directory tree, entrypoints, architecture notes — so an agent onboards in one read instead of thirty.
npx skills add cathrynlavery/repo-atlas
Points an agent at your own statements to audit subscriptions and route spend to the right card. Nothing uploaded, no bank connection.
npx skills add cathrynlavery/spend-optimizer
Systematic diagnosis through progressive disclosure: a tiny router that pulls in one of seven frameworks on demand instead of loading all of them into context.
npx skills add cathrynlavery/problem-solver-skill
The earlier, broader Shopify agent: Hydrogen, Functions, and B2B coverage in one package. Superseded for most uses by the Shopify Developer Skill.
Shopify expertise for Claude Code that checks the current API before writing against it, because Shopify's APIs move and confident deprecated code is worse than none.
npx skills add cathrynlavery/shopify-developer-skill
Setup and operations for running always-on agents across more than one Mac without them tripping over each other.
npx skills add cathrynlavery/mac-agent-ops
Activity recaps for Wispr Flow voice dictation — what you actually dictated, when, and how much of it there was.
npx skills add cathrynlavery/wispr-flow-recap
Reference guides
Cheat sheets, setup notes, and practical guides that are meant to be reused, not just read once.
The first four apps I reinstalled after my laptop died, plus three more that help me capture, remember, and share my work.
Stance prompts that make Fable 5 audit your codebase like a day-one CTO, an activist investor, or a departing cofounder. Copy, paste, run.
Stance prompts that run on your business, calendar, inbox, and family schedule. Hand Fable 5 an identity with something to lose and get owner answers.
Screenpipe records your screen and audio continuously, stores everything locally, and makes it all searchable with AI. Two weeks in, I'm not sure how I worked without it.
g-brain (gbrain) is Garry Tan's open-source knowledge layer for AI agents. A founder's plain-English breakdown of how it works with OpenClaw and gstack.
Every Codex CLI command, safety setting, model pick, and prompting pattern I actually use now that GPT-5.4 is here.
The Claude Code commands, permission modes, model picks, and prompt patterns that actually matter in April 2026.