What is bcm-skills?

This repository contains two skills for Claude Code that help a company build and use a Business Capability Map (BCM). A BCM is a structured chart that lists what a company actually does at a functional level, grouped into categories, so leaders can make decisions about where to invest in technology or AI. The method here was distilled from a Gartner webinar on the subject. The first skill, bcm-creator, walks through building the map. It asks clarifying questions about the company and its goals, researches the business from public sources, and derives capabilities from the ground up rather than forcing the company into a preset template. It then adds color coding (called pace layers) and markers showing where AI applies, producing an HTML summary table and a raster image of the finished map. If you do not have an AI image generation tool set up, the skill writes a ready-to-paste image prompt instead, so the output still works without extra dependencies. The second skill, ai-opportunity-radar, takes the finished map and scores potential AI projects against it. Each use case gets placed on a chart with two axes: how much value it could create and how feasible it is to build. The output also includes a readiness gap chart across seven dimensions, helping a team see not just which AI projects look attractive but which ones the organization is actually prepared to pursue. Both skills install by copying their folders into a specific directory that Claude Code checks automatically. They require only standard Python, with no external libraries. Sibling skills mentioned in the documentation are optional conveniences, and the skills fall back gracefully if those are not present. The design philosophy emphasizes deriving capabilities from real operational data rather than headlines, keeping the stable capability list separate from the strategic overlays that change more frequently, and maintaining unit purity in the map (capabilities only, not outcomes or processes mixed in). The repository includes worked examples for a TSMC case, with input files and a rendered image, so you can see what the finished outputs look like before generating your own.