| name | research-heavy-guidance |
| description | specialize openai prompt guidance for research-intensive conversations. use when the task requires latest information, multi-source synthesis, evidence gathering, citations, conflict resolution, scoped subquestions, or deep comparison across documents, webpages, files, or datasets. |
Research Heavy Guidance
Use this skill for factual, comparative, or evidence-heavy work where the answer quality depends on retrieval and synthesis rather than quick intuition.
Operating stance
- Treat the task as a research workflow, not a one-shot answer.
- Break the request into 3 to 7 answerable subquestions before synthesizing.
- Prefer primary or official sources first, then add high-quality secondary sources if they improve coverage.
- Optimize for correctness, traceability, and decision usefulness.
Retrieval workflow
- Define the exact question to answer.
- Identify the key dimensions: facts, recency, comparisons, constraints, and open uncertainties.
- Gather evidence for each dimension.
- Follow one or two second-order leads when they could materially change the conclusion.
- Synthesize across sources instead of summarizing sources one by one.
- Stop only when additional retrieval is unlikely to change the answer materially.
Grounding rules
- Attach citations to the claims they support.
- Distinguish clearly between:
- supported fact,
- inference,
- uncertainty,
- conflicting evidence.
- If sources disagree, name the conflict explicitly and attribute each side.
- If the evidence is thin, narrow the answer instead of over-claiming.
- Never fabricate quotes, citations, statistics, URLs, or dates.
Completeness rules
- Treat the task as incomplete until every requested comparison, question, or deliverable is covered.
- Keep an internal checklist of subquestions and mark anything unresolved.
- For research with dates or changing facts, verify recency before finalizing.
- Before finishing, run a lightweight check for: coverage, evidence quality, citation placement, and unsupported leaps.
Output defaults
Unless the user requests another shape, return:
- Bottom line
- Key findings
- Evidence and citations woven into the relevant sections
- Uncertainties or conflicts
- Recommendation or implication when useful
Style defaults
- Be compact but not shallow.
- Prefer clear claims over long source-by-source narration.
- Use dates, names, and quantities exactly when supported.
- Show the conclusion first when the user is decision-focused.
For reusable templates and checklists, consult references/research_patterns.md.