User Prompt

Analyze the earnings-call transcript below for the company {company}.

TASK
Extract structured intelligence with these sections:

1. management_tone
   - classification: one of "optimistic", "neutral", "cautiously optimistic",
     "cautiously pessimistic", "pessimistic".
   - confidence: a number between 0 and 1 reflecting how clearly the transcript
     supports the classification.
   - evidence: 3 to 8 VERBATIM quotes from the transcript that justify the
     classification. Copy them exactly.

2. key_takeaways: 3 to 6 of the most material business insights (profitability,
   revenue trends, credit quality, portfolio growth, efficiency, capital, etc.).

3. guidance: explicit forward-looking statements / forecasts / targets the
   management made. Quote or closely reference the transcript.

4. guidance_changes: how guidance and themes changed VERSUS THE PRIOR QUARTER.
   {guidance_changes_instruction}

5. analyst_questions: the TOP 3 most critical questions asked by analysts and how
   management answered. For each: question (paraphrase ok), response_summary, and
   response_quality ("High", "Medium", or "Low") judging how directly and
   completely management answered. Base this ONLY on the analyst-questions block
   and the transcript answers below.

6. red_flags: linguistic signals of hesitation, topic-changing, or evasion. For
   each, provide a "quote" that is a VERBATIM excerpt from the transcript and a
   "reason" explaining why it is a red flag. If there are none, return an empty
   list.

7. surprise_score: an object with "score" (integer 1-10) indicating how much of
   what was said likely was NOT in pre-call market consensus, and a
   "justification" string. Consider the whole transcript, not just the opening.

OUTPUT CONTRACT (return ONLY this JSON object, no fences, no extra text):
{{
  "company": "{company}",
  "management_tone": {{
    "classification": "string",
    "confidence": 0.0,
    "evidence": ["verbatim quote", "..."]
  }},
  "key_takeaways": ["string", "..."],
  "guidance": ["string", "..."],
  "guidance_changes": [
    {{"change": "string", "impact": "string"}}
  ],
  "analyst_questions": [
    {{"question": "string", "response_summary": "string", "response_quality": "High|Medium|Low"}}
  ],
  "red_flags": [
    {{"quote": "verbatim excerpt", "reason": "string"}}
  ],
  "surprise_score": {{"score": 0, "justification": "string"}}
}}

ANALYST QUESTIONS (use these to fill analyst_questions; if empty, infer from the
Q&A inside the transcript):
{analyst_questions}

{prior_quarter_block}

TRANSCRIPT (current quarter - your single source of truth):
{transcript}