We start with an empty folder and turn it into the working home for the consultancy: the prospect's brief and call transcript, past proposals, and the services and pricing doc, all dragged in as plain files. One prompt gives the repo an AGENTS.md so any agent knows what it is and how it's organized. Then we build the first skill. Using bm-skill-builder (a free tool of mine that shows up all through the course), we get a proposal-writing skill that declares what it needs, reads the client folder, uses only real prices, and drafts from a template mined out of a past proposal. It's two markdown files, and it works.
What's covered:
- Organizing the source material: client folders, past proposals, reference docs
- Prompting for an AGENTS.md and what it does for the repo
- Why the repo works in any AI harness
- Building the write-proposal skill with bm-skill-builder
- What a skill is at its simplest: SKILL.md plus a template
- Declared inputs, a template mined from real past work, and where the draft gets saved