Prediction Report Writer

Prediction Report Writer Skill

Write a clear, actionable prediction report from the planned answer strategy and ResearchNotes artifacts. This skill adapts prediction-system synthesis patterns for Deep Researcher Agent's broader research tasks.

Use this skill for probabilistic predictions, point forecasts, ranges, price targets, threshold forecasts, event likelihoods, and forecast-style decision reports. This skill is for final synthesis only; do not perform new research.

Use this skill instead of long-form-report-writer whenever answer_strategy.answer_type is prediction or the user asks for a forecast, stock price, price target, probability, odds, expected value, or next-period value. Only combine it with long-form guidance when the user explicitly asks for a long-form prediction report.

Required Input Review

Before writing, read:

1. /shared/plan.json
2. Every ResearchNotes JSON file under /shared/ that supports the plan
3. The compact result of get_verified_sources

Use think to build a prediction synthesis outline before drafting:

• Identify the original prediction question.
• Identify the target variable: probability, price, value, date, winner, threshold bucket, yes/no outcome, ranking, or option selection.
• Extract all anchors from the research notes, such as latest known values, base rates, market prices, expert forecasts, recent trend data, and official measurements.
• Extract directional evidence supporting the prediction.
• Extract directional evidence opposing the prediction.
• Use each note's ResearchNotes.evidence_judgment when present to prioritize notes: high-score/high-confidence notes should anchor the forecast, medium notes can support or nuance it, and low-score or low-confidence notes should mainly inform gaps, caveats, conflicts, or weak-evidence warnings.
• List major uncertainties that could change the forecast.
• Decide the final forecast before drafting, then write the report around it.

Forecast Discipline

• Commit to a view. Do not say only that the outcome is uncertain.
• Use the strongest available anchor as the starting point:
  • For probabilities: use prediction markets, base rates, polls, odds, expert consensus, or historical frequency when available.
  • For stock/market/economic values: use the latest known value and recent trend data before applying an adjustment.
  • For threshold or multiple-choice predictions: identify the forecasted value first, then select the matching option or bucket.
  • For ranked or categorical outcomes: rank candidates using evidence strength and recent continuity.
• If research notes include an ensemble result, aggregated probability, market consensus, or model estimate, use it as the starting point. Adjust only when the cited evidence strongly warrants it, and explain the adjustment.
• If there is no explicit ensemble, synthesize from the research evidence and say what anchor or base rate you used.
• Express uncertainty through a confidence interval, forecast range, or scenario probabilities rather than vague hedging.
• Avoid language like "it is hard to say" unless immediately followed by a concrete best estimate and uncertainty range.

Required Report Content

The final Markdown answer should include these elements unless the user's requested shape is explicitly shorter:

1. Direct Forecast

   • Put the final probability, value, range, selected option, or prediction at the top.
   • For probabilistic questions, include a probability and confidence interval such as 42% (confidence interval: 35%-55%).
   • For numeric forecasts, include a point estimate and plausible range.
   • For multiple-choice or threshold questions, state the selected option and the underlying forecast value when possible.

2. Executive Summary

   • One concise paragraph explaining the forecast and why.

3. Supporting Factors

   • Three to five cited bullet points or short paragraphs with evidence that supports the forecast.

4. Opposing Factors

   • Two to four cited bullet points or short paragraphs with evidence against the forecast or reasons the forecast could be too high/low.

5. Key Uncertainties

   • Two to four major unknowns that could change the outcome.

6. Recommendations or Implications

   • Two to four actionable recommendations, implications, monitoring points, or decision considerations for someone acting on the forecast.

7. Methodology Note

   • Briefly explain how the forecast was synthesized from evidence. Do not mention agents, prompts, tools, or internal files.

8. Sources

   • Include a compact sources section with every cited source.

Output Shape Guidance

Use clear Markdown. A good default structure is:

# [Prediction Title]

**Forecast:** [direct probability/value/selection]
**Confidence interval / range:** [range]
**Confidence:** [low/medium/high]

## Executive Summary
...

## Supporting Factors
- ...

## Opposing Factors
- ...

## Key Uncertainties
- ...

## Recommendations / Implications
- ...

## Methodology Note
...

## Sources
[1] ...

For a very short requested answer, keep the direct forecast first, then include only the minimum rationale and sources needed to make it trustworthy.

For table-oriented forecast requests, include a table with scenario, forecast, probability or range, evidence, and caveats.