fix(clarify): enhance interactive questioning process with vscode/askQuestions. Fix #1657.#1962
fix(clarify): enhance interactive questioning process with vscode/askQuestions. Fix #1657.#1962fdcastel wants to merge 2 commits intogithub:mainfrom
vscode/askQuestions. Fix #1657.#1962Conversation
There was a problem hiding this comment.
Pull request overview
Updates the /speckit.clarify command template to prefer VS Code’s interactive question UI (vscode/askQuestions) when available, aiming to reduce per-question chat turn overhead in interactive clarification.
Changes:
- Adds guidance to use
vscode/askQuestions(or similar “ask user” tools) during the clarification questioning loop. - Adjusts multiple-choice rendering rules to only show the Markdown options table when the tool is not available.
- Minor rewording/reshaping of the sequential questioning instructions.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
- Allow batching all queued questions in a single vscode/askQuestions call to reduce round-trips (github#1657) - Explicitly include option lists and recommendations in askQuestions prompt - Restore 1-2 sentence limit on recommendation reasoning - Fix sub-bullet formatting for reply instructions (lines 124, 129)
|
I just tested the updated version (from 8376bbd) in a real-world scenario, and it worked perfectly. All questions were asked in a single call, without any repeated turns. |
|
This PR looks good for the use of clarify if you do not assume the user can paste in a long list of questions it wants the clarify agent to iterate over. And then there is also the matter of creating a conditional in a prompt file just for one particular agent. Because of these 2 reasons we cannot take this PR as it stands. We would recommend you publish this as your own preset for the community to try out it out. See https://github.com/github/spec-kit/tree/main/presets for more information. |
Description
This PR simply applies the @stenyin solution to address the issue described in #1657.
I’ve been using this patch successfully for several days now, and it works flawlessly.
Testing
uv run specify --helpuv sync && uv run pytestAI Disclosure
UPDATE: Used GitHub Copilot (Claude Opus 4.6 model) to fix the Copilot AI initial review.