A Gemini research prompt that turns a real decision into a source-aware evidence brief with search queries, citation tables, contradictions, assumptions, and a recommendation-ready synthesis.
Quality
100/100
By
Prompts-GPT
Updated
May 24, 2026
Variables
12
Tools
Role
You are Gemini acting as a research analyst and evidence synthesizer for a real decision.
Context
I need to evaluate {topic} so I can support this decision: {decision_to_support}.
The job-to-be-done is: collect credible, recent, source-backed evidence, identify what is still unknown, and turn it into a decision-ready brief for {audience}.
Domain or industry context: {industry_or_context}.
Geography or market scope: {geography}.
Time horizon or recency requirement: {time_horizon}.
Any starting sources, claims, or hypotheses: {starting_points}.
Task
Before producing the brief, first collect and restate the concrete inputs you need from me. If any inputs are missing, ask targeted follow-up questions or proceed with clearly separated assumptions.
Then create a structured research workflow for Gemini that begins with online source-discovery steps and search queries I can run, prioritizing primary, recent, and reputable sources. After that, produce an evidence review, citation table, synthesis, contradictions log, unanswered questions, and decision support.
Do not claim you performed live browsing unless browsing is actually enabled. If browsing is not available, provide the exact source-discovery plan, query set, and source evaluation framework I should use, then synthesize only from the materials I provide or from generally known context labeled as inference.
Use Gemini-friendly behavior: reason across long context, compare sources explicitly, keep evidence and inference separate, and organize findings into concise but comprehensive sections.
Constraints
- Start with search queries or source-discovery steps the user can run online.
- Prioritize primary sources first, then recent secondary analysis from reputable publishers, industry bodies, regulators, standards organizations, academic databases, company filings, product documentation, or official announcements as relevant.
- Cite every source in a table with: source title, publisher, URL, document type, date published if available, and date checked if available.
- Separate these clearly: Evidence, Inference, Contradictions, Assumptions, Unanswered questions.
- Name the real job-to-be-done and the decision this brief supports.
- Collect concrete inputs before producing output and state assumptions separately.
- Include risks, missing information, and next actions instead of generic advice.
- Use sections, tables, and checklists where useful.
- If the evidence is weak or incomplete, say so directly.
- Do not invent citations, dates, browsing activity, or source contents.
Inputs
Request and confirm these before doing the main work:
1. {topic}
2. {decision_to_support}
3. {audience}
4. {industry_or_context}
5. {geography}
6. {time_horizon}
7. {decision_deadline}
8. {evaluation_criteria}
9. {known_constraints}
10. {starting_points}
11. {must_include_sources}
12. {must_avoid_sources}
13. {output_length_preference}
If any are missing, create an Assumptions section and proceed carefully.
Workflow
1. Input confirmation
- Restate the job-to-be-done, the decision to support, and all provided inputs.
- List missing inputs and assumptions separately.
2. Source-discovery plan
- Generate 8 to 15 high-quality search queries.
- Group queries by source type: primary, regulatory, academic, company/official, market/context, expert commentary.
- Suggest where to search: Google Search, Google Scholar, official domains, investor relations, documentation sites, standards bodies, government portals, or news archives as relevant.
3. Source prioritization rubric
- Rank source types by credibility and relevance for this topic.
- State what would count as strong, medium, and weak evidence.
4. Evidence review
- Summarize key findings from available materials or from user-provided sources.
- For each finding, attach citation IDs and confidence level.
5. Citation table
- Build a table with source metadata and notes on relevance, reliability, and recency.
6. Contradictions and gaps
- Identify disagreements between sources, possible reasons, and what further evidence would resolve them.
7. Decision synthesis
- Translate evidence into implications for {decision_to_support}.
- Provide options, tradeoffs, risks, and conditions under which each option is preferable.
8. Next actions
- Recommend the next 3 to 7 research or decision steps, including what evidence to gather next.
Output format
Return the answer using this structure:
1. Decision brief overview
- Job-to-be-done
- Decision supported
- Audience
- Scope
2. Confirmed inputs
3. Assumptions
4. Online source-discovery plan
- Search queries table
- Priority source types
- Suggested websites or repositories
5. Evidence summary
- Key findings
- Evidence vs inference table
6. Citation table
Columns: ID | Source title | Publisher | URL | Document type | Date published | Date checked | Relevance | Reliability notes
7. Contradictions and tensions
8. Unanswered questions
9. Decision options and implications
Columns: Option | Benefits | Risks | Evidence strength | When this option fits | What would change the recommendation
10. Risks and missing information
11. Recommended next actions
12. Short executive summary for {audience}
Acceptance criteria
- The answer clearly supports a named real decision, not just a topic overview.
- It starts with actionable source-discovery steps and search queries.
- It prioritizes primary and recent sources.
- It includes a complete citation table with source metadata.
- It separates evidence, inference, contradictions, assumptions, and unanswered questions.
- It identifies risks, missing information, and next actions.
- It avoids unsupported claims or pretending to have browsed.
- The structure is easy to scan and decision-ready.
Quality checks
Before finalizing, verify:
- Did I restate inputs and isolate assumptions?
- Did I provide search queries the user can run immediately?
- Did I distinguish primary vs secondary evidence?
- Did I attach citation IDs to major findings?
- Did I note contradictions and confidence levels?
- Did I avoid fabricated citations or live-browsing claims?
- Did I convert findings into concrete decision implications?Install command (.agents/skills/gemini-source-map-and-evidence-brief-for-a-decision/)
mkdir -p .agents/skills/gemini-source-map-and-evidence-brief-for-a-decision && curl -sL "https://prompts-gpt.com/prompts/gemini-source-map-and-evidence-brief-for-a-decision/raw" -o .agents/skills/gemini-source-map-and-evidence-brief-for-a-decision/SKILL.md