User Story

As a solo founder running ApexYard, I want a lightweight 5-question idea-validation skill I can invoke either standalone (/validate-idea IDEA-NNN) or as an offered follow-up step inside /idea and /handover, so that I have a 10-minute pre-spec gate that catches "this idea isn't worth a PRD" without needing event storming, Wardley mapping, or any heavyweight strategy ceremony.

Context

/idea captures an idea to the backlog quickly; /write-spec produces a PRD. Between them sits a quiet gap — "is this idea worth speccing?". Today the implicit answer is "yes, just spec it" — but that's an expensive default. Most ideas SHOULD die before the PRD round.

/handover ingests external repos. Some are dormant or contractor-orphaned, and the question "should we keep pursuing this product at all?" is exactly the validation question — but it's hidden inside the integration plan and rarely asked explicitly.

A lightweight /validate-idea skill — five questions, ten minutes, one-page output — fills the gap. It's invokable standalone, and offered as a follow-up step inside both /idea (after capture) and /handover (after the integration plan).

Acceptance Criteria

• New skill at .claude/skills/validate-idea/SKILL.md with argument-hint: "<IDEA-NNN | project-name | free-form description>".
• Asks five questions, one at a time (not batched, per the existing skill convention):
  1. Target user — Who specifically is this for? (Not "everyone")
  2. Current alternative — What do they do today instead? (If "nothing", probe deeper — is this really a problem?)
  3. Smallest version — What's the smallest thing that would prove the value? (1-2 days of work)
  4. Kill criteria — What would prove this is wrong? (Concrete signal that ends the experiment)
  5. Build, buy, or rent — Is this commodity (use OSS), differentiating (build), or somewhere we should rent (SaaS)?
• Output: a one-page markdown doc at:
  • projects/<project>/validation/IDEA-NNN-validation.md if invoked on an IDEA-NNN from the backlog
  • projects/<project>/validation/handover-validation.md if invoked from /handover
  • projects/_inbox/validation/<slug>.md for free-form invocations not tied to a registered project
• Output ends with a one-line verdict: green (proceed to /write-spec), yellow (revise + revalidate), or red (park / archive — append to projects/ideas-backlog.md with status WONTDO).
• On red, automatically update the IDEA-NNN row in the backlog to WONTDO (if the input was an IDEA-NNN). On green, leave the status as NEW so the next step (/write-spec) makes the transition.
• Updated /idea skill spec: at the end, after the IDEA-NNN is captured and the optional GitHub Issue is offered, add a third optional step: "Validate now? (run /validate-idea IDEA-NNN — y/n, default n)". Default-no respects the lightweight capture intent.
• Updated /handover skill spec: at the end of the integration-plan emit, add an optional step: "This project looks dormant / contractor-orphaned — run /validate-idea to confirm it's still worth investing in? (y/n, default n)". The "looks dormant" trigger is heuristic: last commit > 90 days ago, no recent PRs, no issues with activity.
• Updated CLAUDE.md skills table: /validate-idea row added.

Design Notes

• Skill structure: match the existing /idea shape (front-matter, numbered Process steps, Rules section). No tests — markdown spec.
• Question order matters: 1 → 2 sets up "is there a real gap?". 3 sets up "is the smallest test cheap?". 4 forces falsification thinking (most ideas can't articulate kill criteria — that itself is signal). 5 forces the build-or-rent call early.
• No scoring rubric. Verdict is a judgement call by the user reading their own answers; the skill nudges with prompts but doesn't auto-grade. (Tried scoring rubrics; they're either too coarse to be useful or too detailed to fit on one page.)
• Integration with /idea: the offer is optional and default-no. The capture flow stays light. Most ideas won't get validated immediately; they get triaged later in batch.
• Integration with /handover: the offer is conditional on the dormancy heuristic and default-no. Healthy active projects don't get the prompt; dormant ones do.

Out of Scope

• Wardley mapping (heavy methodology, wrong audience).
• Event storming (heavy methodology, wrong audience).
• Automated competitive research / market sizing — too specific to product class; leave to the user's own desk research.
• Linking to external validation tools (Mom Test, JTBD canvas, etc.) — leave as see-also bullets in the skill, not framework integration.

Effort Estimate

Small. ~150 lines of SKILL.md + ~10 lines each in /idea and /handover for the optional follow-up step + 1 line in CLAUDE.md.

Glossary