Engineer at the intersection of life sciences and software.
Python CLI and OpenClaw skill that monitors Recreation.gov campsite and wilderness-permit availability, pairs it with Open-Meteo weather, and pipes results to Telegram. Three modes — weekend recon, free-text search, and a cron watch that notifies only on actual openings. Modular architecture with per-module wiki docs aimed at LLM agents reading the codebase.
Repurposed a 32GB / dedicated-GPU machine into a headless agent host. Skills are markdown briefings the agent loads into context; cron drives the loop; Telegram is the interface. Showcase use case: it runs CampsiteRecon on a daily watch and pings me only when a Point Reyes / Big Sur / Sequoia site actually opens up — silent otherwise.
On why generic LLM output stays generic until you hand the model a one-page skill briefing — and how to write one. Patterns I use across CampsiteRecon, the OpenClaw infra, and day-job pipeline work.