home / skills / poemswe / co-researcher / synthesize

synthesize skill

safe

This skill synthesizes findings from multiple sources into coherent conclusions with quantified uncertainty for informed decision making.

npx playbooks add skill poemswe/co-researcher --skill synthesize

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

1.3 KB

---
name: synthesize
description: Synthesize findings from multiple sources into coherent conclusions with uncertainty quantification.
metadata:
  short-description: Research Synthesis
---
# /synthesize - Research Synthesis

I'll integrate findings from multiple sources into coherent, actionable conclusions.

## What I'll do:

### Pattern Recognition
- Identify convergent findings
- Flag contradictions
- Map evidence strength

### Weight Assignment
- Assess source quality
- Apply evidence hierarchy
- Calculate confidence levels

### Contradiction Resolution
- Identify genuine disagreements
- Assess relative credibility
- Propose resolutions

### Narrative Construction
- Organize thematically
- Build logical flow
- Connect pieces coherently

### Uncertainty Quantification
- Classify confidence levels
- Acknowledge limitations
- Use calibrated language

## What I need:

Provide:
- Multiple sources to synthesize, OR
- Research question with context, OR
- Findings from previous research phases

## Output:

You'll receive:
- Executive summary
- Key findings with confidence levels
- Areas of uncertainty
- Contradictions and interpretations
- Integrated conclusions
- Implications for different audiences
- Research gaps identified

What would you like me to synthesize?
Project: $ARGUMENTS

Overview

This skill synthesizes findings from multiple sources into coherent, actionable conclusions with explicit uncertainty quantification. It highlights convergent evidence, flags contradictions, and produces a structured narrative you can use for decision-making or further research. Outputs emphasize confidence levels, limitations, and clear implications for different audiences.

How this skill works

I ingest provided sources, findings, or a research question and perform pattern recognition to spot agreements and conflicts. I assign weights based on source quality and evidence hierarchy, estimate confidence levels, and construct an integrated narrative that links evidence to conclusions. The result includes an executive summary, key findings with quantified confidence, identified contradictions, and prioritized research gaps.

When to use it

Combining results from multiple studies, reports, or datasets into a single set of conclusions
Producing evidence-based recommendations for policy, product strategy, or clinical decisions
Preparing literature reviews, white papers, or briefings that require calibrated uncertainty
Resolving apparent contradictions across sources before presenting to stakeholders
Identifying where further research or data collection will most reduce uncertainty

Best practices

Provide full sources or clear summaries of findings, including methods and sample sizes when available
State the research question and decision context to focus synthesis priorities
Indicate any known biases or conflicts of interest in sources up front
Request both high-level executive summaries and detailed appendices for technical audiences
Use the confidence levels to guide decisions, not as absolute proof

Example use cases

Synthesize clinical trial outcomes and observational studies to inform treatment guidelines, with confidence tiers for each recommendation
Compare market research reports and user studies to produce a product roadmap and highlight uncertain assumptions
Combine environmental impact assessments and regulatory reports to form policy options and note evidence gaps
Aggregate academic literature for a grant proposal, identifying where new data would most improve certainty

FAQ

How do you quantify confidence?

I weight source quality, study design, sample size, and consistency across findings to assign calibrated confidence labels (e.g., high/medium/low) and, where appropriate, numerical ranges.

Can you handle contradictory evidence?

Yes. I flag genuine disagreements, assess relative credibility, and propose plausible resolutions or recommend targeted data collection to resolve the conflict.