questioning

Original：🇺🇸 English
Translated

One-shot question answering using the research process. Answers inline without generating documents, then offers handoff to brainstorm or research.

5installs
Sourcedesplega-ai/ai-toolbox
Added on

NPX Install

npx skill4agent add desplega-ai/ai-toolbox questioning

Tags

Translated version includes tags in frontmatter
one-shot-question-answeringagent-skillresearch-workflowcodebase-investigationhandoff-mechanism

SKILL.md Content

View Translation Comparison →

Questioning

You are answering a question directly and concisely using the research process. No documents are created by default — the answer is the deliverable.

When to Use

This skill activates when:
  • User invokes 
    /question
    command
  • Another skill references 
    desplega:questioning
  • User asks a focused question that doesn't need a full research document

Philosophy

This is the fastest path from question to answer. Unlike 
/research
(which documents comprehensively) or 
/brainstorm
(which explores interactively), 
/question
gets straight to the point:
  1. Hear the question
  2. Investigate as needed
  3. Answer inline
  4. Offer next steps
No ceremony — no autonomy prompts, no preference setup, no working agreement. Just answer the question.
When you do need to ask the user something (handoff, follow-up choice), use 
AskUserQuestion
per 
desplega:ask-user
conventions.

Process

Step 1: Analyze the Question

Classify the question to determine what investigation is needed:
Question TypeInvestigation NeededExample
Factual/locationalQuick codebase search"Where is the auth middleware defined?"
Conceptual/howRead relevant files"How does the plugin system discover skills?"
Why/root causeDeep read + history"Why does brainstorming default to verbose?"
ComparativeRead multiple areas"What's the difference between research and question skills?"
External/librarycontext7 or web search"How does Bun's SQLite driver handle transactions?"

Step 2: Investigate

Based on the question type, use the appropriate tools. Spawn sub-agents only when needed — many questions can be answered by reading a few files directly.
For codebase questions:
  • Read directly mentioned files first (use Read tool WITHOUT limit/offset)
  • Use codebase-locator agent if you need to find WHERE something lives
  • Use codebase-analyzer agent if you need to understand HOW something works
  • Use codebase-pattern-finder agent if you need examples of a pattern
For library/framework questions:
  • Use context7 MCP to fetch documentation (
    resolve-library-id
    → 
    query-docs
    )
For external/web questions:
  • Use web-search-researcher agent for documentation or examples
Efficiency rule: If you can answer by reading 1-3 files, just read them. Don't spawn sub-agents for simple lookups.

Step 3: Answer

Present the answer as inline text (not a document). Structure it naturally based on the question:
  • Short answers: 1-3 sentences with a file:line reference
  • Medium answers: A paragraph or two with key references
  • Detailed answers: Structured with headings if needed, but keep it focused
Unlike research, you MAY:
  • Suggest improvements if the question implies a problem
  • Perform root cause analysis if the question asks "why"
  • Give opinions when asked ("which approach is better?")
  • Be direct and opinionated rather than exhaustively neutral
Always include:
  • Specific 
    file:line
    references for any claims about the codebase
  • Code snippets when they clarify the answer (keep them short)

Step 4: Handoff

After answering, use AskUserQuestion with:
QuestionOptions
"What would you like to do next?"1. Ask another question, 2. Save this answer to thoughts, 3. Start a brainstorm from this topic (→ 
/brainstorm
), 4. Start research from this topic (→ 
/research
), 5. Done
Based on the answer:
  • Ask another question: Use AskUserQuestion to ask "What's your next question?" and loop back to Step 1
  • Save this answer: Write the Q&A to 
    thoughts/<user>/questions/YYYY-MM-DD-<topic>.md
    using the template at 
    cc-plugin/base/skills/questioning/template.md
    . Path selection: use the user's name if known, fall back to 
    thoughts/shared/questions/
  • Brainstorm: Suggest 
    /brainstorm <topic>
    with the question's topic as context
  • Research: Suggest 
    /research <topic>
    with the question's topic as context
  • Done: No further action

Looping Behavior

When the user selects "Ask another question," the skill loops:
  1. Ask for the next question via AskUserQuestion
  2. Investigate and answer (Steps 1-3)
  3. Present handoff options again (Step 4)
Each iteration is independent — no state accumulates between questions unless the user explicitly connects them.

Learning Capture

OPTIONAL SUB-SKILL: If significant insights, patterns, gotchas, or decisions emerged during this workflow, consider using 
desplega:learning
to capture them via 
/learning capture
. Focus on learnings that would help someone else in a future session.

What This Skill is NOT

  • Not research: No comprehensive document, no frontmatter ceremony, no multi-section output
  • Not brainstorming: No Socratic Q&A loop, no progressive document
  • Not a chatbot: Each question gets proper investigation with codebase evidence, not surface-level responses