text2qa Skill

Extract Q&A pairs and Selection Preferences from the current chat session or any provided text.

Output Format

Produce two clearly separated sections:

Section 1: Q&A Pairs

Identify every implicit or explicit question-answer exchange. Format as:

## Q&A Pairs

**Q1: <question>**
A: <answer>

**Q2: <question>**
A: <answer>
...

Rules:

• Extract both explicit questions (user directly asked something) and implicit questions (Claude provided info that answers an unstated question).
• Rephrase conversational exchanges into clean, standalone Q&A pairs.
• If an answer spans multiple turns, synthesize into one coherent answer.
• Skip small talk or meta-conversation (e.g., "thanks!", "sure!").

Section 2: Selection Preferences

Identify any preferences, constraints, choices, or criteria the user expressed or implied. Format as:

## Selection Preferences

| # | Preference / Constraint | Source (direct/inferred) |
|---|------------------------|--------------------------|
| 1 | <preference>           | direct / inferred        |
| 2 | <preference>           | direct / inferred        |
...

Preference types to look for:

• Format preferences — "I want markdown", "keep it short", "use bullet points"
• Content preferences — "focus on X", "skip Y", "I prefer Z style"
• Tool/approach preferences — "don't use files", "use Python not JS"
• Persona/tone preferences — "be concise", "explain like I'm a beginner"
• Domain/topic constraints — "only for production use", "target audience is X"
• Decisions made — explicit choices the user made during the session

Mark each as direct (user stated it outright) or inferred (implied by behavior/choices).

Workflow

1. Scan the full conversation (or provided text).
2. Identify all Q&A exchanges and stated/implied preferences.
3. Output Section 1 (Q&A Pairs) then Section 2 (Selection Preferences).
4. Optionally offer to export as a .md file if the user might want to save it.

Tips

• For chat sessions: scan from the very first message.
• For documents: treat headings/paragraphs as implicit questions when appropriate.
• If the session is very long, cluster related Q&As under sub-headings by topic.
• If no preferences are found, say so clearly rather than inventing them.
• Always note the total count: "Found X Q&A pairs and Y preferences."