| name | academic-research |
| description | Support rigorous academic research workflows including literature discovery, paper triage, reading notes, evidence synthesis, citation hygiene, research planning, reproducible writing, reviewer responses, and safe handling of original data. Use when Codex is asked to find, evaluate, summarize, compare, cite, organize, or write about scholarly sources, research questions, methods, experiments, datasets, or academic manuscripts. |
Academic Research
Core Workflow
- Clarify the research target: identify the research question, discipline, deliverable, audience, scope, deadline, and required citation style.
- Gather source context: inspect provided papers, notes, BibTeX, Zotero records, project files, or source lists before searching externally. Use current, primary, peer-reviewed sources when recency or precision matters.
- Screen evidence: rank sources by relevance, quality, method fit, venue credibility, recency, citation provenance, and whether they directly support the user's question.
- Extract structured notes: capture each source's claim, method, assumptions, data, result, limitations, and reusable citation metadata.
- Synthesize across sources: group agreements, disagreements, methods, empirical settings, gaps, and open questions. Distinguish evidence from inference and speculation.
- Produce the requested artifact: create the literature matrix, reading notes, synthesis memo, research plan, outline, draft section, citation audit, reviewer response, or reproducibility checklist.
- Preserve provenance: attach citations to claims, record uncertainty, and avoid unsupported generalizations.
Research Standards
- Prefer primary sources over secondary summaries for claims about methods, results, definitions, or chronology.
- Cite precisely enough that the user can trace every substantive claim back to a source.
- Mark unverified claims clearly when source access is incomplete.
- Compare papers on method, assumptions, data, findings, and limitations rather than summarizing each paper in isolation.
- Preserve the user's terminology for project-specific concepts unless a source uses a better-established term.
- For mathematical or technical research, separate theorem statements, assumptions, proof strategy, empirical evidence, and implementation details.
- For empirical research, note sample/data provenance, identification strategy, baselines, metrics, robustness checks, and threats to validity.
- For interdisciplinary work, state which discipline's convention is being followed when terminology or evidence standards differ.
Data And File Safety
- Treat raw, original, source, experimental, log, measurement, and archival data as read-only.
- Never overwrite, filter, normalize, rename, or delete original data files.
- Put derived outputs in clearly labeled new files or folders when file creation is requested.
- When unsure whether a file is original data, ask before modifying it.
- When proposing scripts or analyses, design them to read original inputs and write derived copies.
Artifact Templates
Read references/research-artifacts.md when the user asks for a literature matrix, reading note, synthesis memo, research plan, citation/provenance audit, or reproducibility checklist.
Use the templates flexibly: keep headings that fit the request, remove empty sections, and prefer concise research-useful content over decorative formatting.
Quality Check
Before finalizing research work, verify:
- The answer directly addresses the research question and requested deliverable.
- Each important claim has a source, a stated uncertainty, or a clear label as inference.
- Conflicting evidence is represented fairly.
- Limitations and missing evidence are explicit.
- Citations, filenames, and source identifiers are internally consistent.
- Any proposed data processing preserves original data untouched.
