Why Now Skill — Productivity skill for Claude Code

why-now

A Claude Code skill that runs a short value check *before* any planning or implementation work starts — is this task worth doing, and worth doing now?

It's self-contained: the value gate (§1–8) decides *whether* to build, and §9 decides *how*. §9 prefers to use a fuller design skill (superpowers:brainstorming) when one is available — and will offer to install it once, if it isn't — because that does a noticeably better job than a compact built-in version can. If that's declined or unavailable, it falls back to its own lightweight bounded/architectural design pass, so the skill still works end to end with nothing else installed.

why-now: value gate  →  §9 (design, real skill if possible, else built-in)  →  implementation

What it does

Before touching any real work (a feature, a refactor, a new project, a script), it:

1. States the cost honestly — rough time, rough tokens, who maintains it, how reversible it is.
2. Asks up to three targeted questions (never more) from six possible angles: who's actually hurting, why now, what "done" looks like, the cost of doing nothing, what it displaces, and whether a 10%-effort version already answers the question.
3. Gives one of four verdicts: GO, SHRINK (do the cheap version first), SHELVE (not now, with a return trigger), or DROP.

On GO or SHRINK, it moves into its own design pass: for a small, bounded change it asks a couple of questions and proposes a short design in chat; for something architectural (new project, new subsystem) it asks purpose and constraints, proposes 2–3 approaches, and writes the agreed design to a spec file in the project's own folder. Either way, implementation only starts after an explicit yes.

It skips itself entirely for questions, code search, typo fixes, and live incidents — it never gets in the way of work that's obviously worth doing.

A skip phrase ("I know, just do it") bypasses the whole thing with one line of acknowledgment — no l