What is actually in the SKILL.md
A skill is only as good as its instructions, so we simply describe them here. At the top is a frontmatter with the conditions under which Claude should pick up the skill: on the word backcasting itself and variants such as reverse roadmap and future back planning, but also on sighs such as I don't know how to get there, my goal is too big, and my plan doesn't feel ambitious enough, it feels incremental. Anyone who says something like that to Claude while the skill is loaded gets it automatically. The frontmatter also names version 1.0.0, the MIT licence and the two founders, Robinson and Robèrt.
Then follows the framework in five phases. Phase one defines the future vision, and the file sets the bar explicitly: a good vision is specific, time bound, desirable, not limited by current resources and imaginable. It asks questions about it out loud: what year is it in the vision, what does an ordinary day look like, which figures belong to it, what do you no longer do then, who is on the team and who are the customers, and what is different about the world, your industry or your region.
Phase two describes the present honestly, with no invented starting point: where do you stand now on those same variables, which resources, skills, network and capital are there, which habits, customers, contracts or identity are fixed, and what is the real constraint instead of the self imposed one.
Phase three makes the gap explicit on five axes: financial, operational, people, market and personal. Per axis not only the difference comes on the table, but also the question of which patterns need to be broken instead of optimised.
Phase four is the heart: the backcast in milestones, from the target year via T-1, T-3 and T-5 back to now, exactly the sequence from FIG.02. Every milestone must describe a measurable state instead of an activity, be logically reasoned back to the previous one, and not conflict with physical or fundamental laws.
Phase five closes with the first steps and blockers: three concrete actions for this month, three patterns that need to be broken, three assumptions to test, three early signals that show the vision is becoming more achievable, and what you can already stop doing now because it does not contribute to the goal.
The file closes with five hard rules, which you see in action in FIG.04 below, and a source list of four titles: Robinson from 1982, Robèrt from 1989, Holmberg and Robèrt from 2000 and Dreborg from 1996. If you want to learn to build a file like this yourself, the full explanation is in writing a SKILL.md.