Summary
Builder and creator Brian Casel walks through four production agent skills he uses daily to run the marketing side of his business via OpenClaw running on a Mac Mini. Rather than showcasing AI demos, the video reveals the specific structure behind each skill: the file layout, training references, scheduling logic, and tool integrations that make each workflow reliable enough to run unattended.
The first skill — a “radar scan” — runs at 4 a.m. via a Telegram-connected agent named Veil, which monitors curated feeds and surfaces relevant industry content as a markdown report waiting in Telegram each morning. The second skill handles writing and sending Casel’s free weekly newsletter, the Builder Briefing. The third is a brand illustration workflow that uses a structured interviewing process to develop a visual concept, then calls the Google image generation API to produce on-brand assets — referencing a custom color palette and visual world baked into the skill’s reference files. Each skill lives in a folder with a main skill file and supporting references the agent consults during execution.
Casel also introduces two custom tools he built: Brainown, a markdown reader/writer for passing files between agents and humans, and Spark Drop, a tool for storing and retrieving personal interest and voice training data. The core argument — that any repeatable business process can be turned into an agent skill — applies well beyond marketing, and the video provides a practical mental model for structuring agent work in OpenClaw or similar frameworks.
📺 Source: Brian Casel · Published April 06, 2026
🏷️ Format: Workflow Case Study
