AskUserQuestion: raise maxItems cap from 4 (or make configurable) for multi-question methodology gates

Resolved 💬 1 comment Opened May 23, 2026 by jasonlanecg Closed Jun 23, 2026

Summary

The built-in AskUserQuestion tool enforces maxItems: 4 on its questions array. For methodology / discovery workflows that need to present 5-8 related questions in a single batch (e.g. acceptance-criteria clarification skills, multi-axis design reviews), this forces awkward chaining of 4 + 1 calls with two separate UI prompts.

Concrete use case

Building an RPI-SDD /clarify skill that does a 9-taxonomy ambiguity scan on a feature spec. The skill's own body says "Maximum 5 questions per run" — picked because 5 is the sweet spot for an operator to triage in one sitting. The tool's hard cap at 4 forces:

Call 1: AskUserQuestion with Q1..Q4
Call 2: AskUserQuestion with Q5 ("deferred from cap above")

The operator sees two separate UI prompts for what's logically one decision batch, and the second prompt has to carry context ("Q5 of 5, deferred from previous").

Proposed change

Either (in order of preference):

  1. Raise maxItems to 8 or 10 — covers all common methodology-gate use cases without becoming unwieldy.
  2. Make maxItems configurable via settings.json (e.g. "toolDefaults.AskUserQuestion.maxItems": 8), defaulting to 4.
  3. Document the chained-call pattern as the recommended approach so skill authors don't try to fit > 4 questions into one call.

Workaround in use today

Splitting into chained calls with explicit pass numbering in question headers. Functional but produces a worse operator experience than a single combined prompt.

Context

This came up while running an RPI-SDD /clarify gate on a 15-AC feature spec. 5 high-impact ambiguities surfaced; we wanted to present them in one batch with the same priority framing. The 4-cap forced "Q5 deferred from cap" framing that the operator rightly questioned.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗