# Research Synthesis Brief

> Synthesize multiple sources, reports, notes, articles, datasets, interviews, or research excerpts into claims, evidence, caveats, contradictions, confidence levels, and recommended actions. Use when Codex needs to create source-grounded research briefs, executive summaries, evidence reviews, or decision support.

- Skill: `clarentcinematics/research-synthesis-brief` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add clarentcinematics/research-synthesis-brief`
- Raw SKILL.md: https://api.skillmd.com/api/skills/clarentcinematics/research-synthesis-brief/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: ClarentCinematics (https://skillmd.com/u/clarentcinematics)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/clarentcinematics/research-synthesis-brief

---


# Research Synthesis Brief

## Workflow

1. Identify research question, audience, decision context, source set, and time sensitivity.
2. Extract claims, evidence, methods, dates, limitations, contradictions, and source quality signals.
3. Separate what is known, what is inferred, what conflicts, and what remains unknown.
4. Synthesize findings into decision-relevant implications rather than source-by-source summaries.
5. Produce a brief with confidence levels, caveats, and recommended next actions.

## Output Standard

Use this structure by default:

- **Research Question**: what the brief answers.
- **Bottom Line**: concise answer with confidence level.
- **Key Findings**: claims with evidence and caveats.
- **Evidence Table**: source, claim supported, strength, limitation.
- **Contradictions / Gaps**: conflicts, missing data, and uncertainty.
- **Decision Implications**: what the findings mean for action.
- **Recommended Next Steps**: follow-up research, validation, or decision path.

## Rules

- Do not cite or rely on sources not provided unless the user asks for external research.
- Label confidence and uncertainty explicitly.
- Do not flatten conflicting evidence into a false consensus.
- Keep source-grounding clear enough for a human reviewer to audit.

## References

Read `references/evidence-rubric.md` when the user asks for evidence grading, confidence scoring, contradiction handling, or decision implications.