What task are you trying to do?
Use PawWork's structured question tool as a low-friction way for the agent to clarify user intent, present real trade-offs, and collect decisions before continuing work. This matters when the next action depends on user preference, scope, risk tolerance, or implementation direction.
What do you do today?
The question tool exists, but the experience is still too compressed. Longer questions or trade-off explanations can appear as one dense block because the question text is rendered as plain text without preserving line breaks or formatting. Multi-select is supported in the protocol and UI, but it is easy to miss unless the agent explicitly sets the right field and the prompt makes that mode clear. The agent also does not call the tool proactively enough, so ambiguity often falls back to ordinary chat instead of a structured decision point.
What would a good result look like?
The agent should use the question tool more actively when a user decision would materially affect the next step, while still avoiding trivial interruptions. The UI should make structured questions easier to scan: preserve intentional line breaks at minimum, make single-select versus multi-select obvious, and consider a separate optional body or description area for context, trade-offs, and longer explanations instead of forcing everything into the question title. Preset options should remain concise, with descriptions carrying the details, and custom answers should remain available.
Which audience does this matter to most?
Both
Extra context
Comparable tools already treat structured questions as a dedicated elicitation surface. Claude Code's AskUserQuestion supports concise options, descriptions, multiSelect, Other, and optional previews for richer comparisons. Kimi Code documents single-select, multi-select, custom text input, multi-question tabs, keyboard navigation, and restoring previous selections. Gemini CLI's ask_user supports choice, text, yes/no, multiSelect, placeholders, and renders question text through a Markdown-capable dialog. For PawWork, a pragmatic first slice is probably: preserve line breaks in question text, improve the visible single-select or multi-select hint, and tune agent instructions so it chooses the tool for meaningful ambiguity and approach selection.
What task are you trying to do?
Use PawWork's structured question tool as a low-friction way for the agent to clarify user intent, present real trade-offs, and collect decisions before continuing work. This matters when the next action depends on user preference, scope, risk tolerance, or implementation direction.
What do you do today?
The question tool exists, but the experience is still too compressed. Longer questions or trade-off explanations can appear as one dense block because the question text is rendered as plain text without preserving line breaks or formatting. Multi-select is supported in the protocol and UI, but it is easy to miss unless the agent explicitly sets the right field and the prompt makes that mode clear. The agent also does not call the tool proactively enough, so ambiguity often falls back to ordinary chat instead of a structured decision point.
What would a good result look like?
The agent should use the question tool more actively when a user decision would materially affect the next step, while still avoiding trivial interruptions. The UI should make structured questions easier to scan: preserve intentional line breaks at minimum, make single-select versus multi-select obvious, and consider a separate optional body or description area for context, trade-offs, and longer explanations instead of forcing everything into the question title. Preset options should remain concise, with descriptions carrying the details, and custom answers should remain available.
Which audience does this matter to most?
Both
Extra context
Comparable tools already treat structured questions as a dedicated elicitation surface. Claude Code's AskUserQuestion supports concise options, descriptions, multiSelect, Other, and optional previews for richer comparisons. Kimi Code documents single-select, multi-select, custom text input, multi-question tabs, keyboard navigation, and restoring previous selections. Gemini CLI's ask_user supports choice, text, yes/no, multiSelect, placeholders, and renders question text through a Markdown-capable dialog. For PawWork, a pragmatic first slice is probably: preserve line breaks in question text, improve the visible single-select or multi-select hint, and tune agent instructions so it chooses the tool for meaningful ambiguity and approach selection.