Acumen: Skill for Claude Code

MANDATORY PREPARATION

1. Invoke /product-thinking — it contains the Context Gathering Protocol and the AI Slop Test. Follow the protocol before proceeding.
2. Diverge first — always. An increment never scopes a path that hasn't been diverged. If a /workshop report already exists for this problem and a path was chosen, read it and proceed. If not, run /workshop now, produce the report, and converge on a single path before writing a line of the increment. The explored options live in the workshop report; the increment scopes only the chosen path and does not re-litigate the alternatives. Scale the divergence to the size — a Small fix gets a quick fork exploration, not a full workshop — but it happens.

Mindset

An increment is a commitment to ship a complete user outcome. Not a feature request. Not a brainstorm. Not a backlog item with aspirations. It is the document that says: "This is what we are building, why, for whom, and how we know it worked." If you cannot fill every section with specific, grounded information, the increment is not ready to build.

This skill scopes one path — the shape that won its /workshop divergence. The alternatives were explored in the workshop report and deliberately set aside; the increment commits to a single shape and doesn't re-open the fork. If the path isn't chosen yet, you're not ready to scope — go diverge first. Don't scope two increments in parallel.

A spec is a thinking tool, not a documentation exercise. The competitive context should shape the scope. The persona reactions should challenge the solution. The user signal should validate the problem. If these inputs don't change what you write, you are not using them.

Scope by outcome, not by time. "What can we build in two weeks?" produces mediocre features. "What is the minimum that changes user behavior?" produces focused ones. Cut features, not corners.

Length discipline

A reader should absorb the whole increment in a few minutes. Terseness is a feature, not a compromise.