# Apply Imrad

> Decide whether IMRaD fits, review an IMRaD draft, transform mixed material into IMRaD, or draft a new evidence-traceable IMRaD output without fabricating results.

- Skill: `narumiruna/apply-imrad` (Agent Skill)
- Install (CLI): `npx skillmds add narumiruna/apply-imrad`
- Raw SKILL.md: https://api.skillmd.com/api/skills/narumiruna/apply-imrad/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: narumiruna (https://skillmd.com/u/narumiruna)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/narumiruna/apply-imrad

---


# Applying IMRaD

Use IMRaD only when a clear question, a real or responsibly inferable method, results or explicitly non-empirical outputs, and interpretation all exist. Do not force it onto simple explanations, tutorials, brainstorming, or style-only edits.

## Select One Mode

| Mode | Trigger | Deliverable |
| --- | --- | --- |
| Detector | applicability, readiness, or section-gap question | fit verdict and gap analysis, not a full article |
| Reviewer | existing IMRaD draft needs audit | prioritized structural/evidence findings, not a rewrite unless asked |
| Recomposer | notes or mixed material must become IMRaD | full transformed draft |
| Writer | new research-style output is requested | full new draft |

Infer the mode when clear. Ask only when the choice materially changes the deliverable.

## Evidence Rules

- Never fabricate data, citations, experiments, measurements, or observations.
- Label non-empirical results as `inferred`, `hypothetical`, `expected`, or `simulated`.
- Put material assumptions in Methods and trace every major Results or Discussion claim to source evidence or an explicit assumption.
- Keep reporting separate from interpretation. Do not hide missing evidence behind fluent prose.
- Add `Abstract` or `Conclusion` only when requested.

## Section Boundaries

1. **Introduction:** context, problem/question, objective or hypothesis. No methods, findings, or conclusions.
2. **Methods:** approach, process, tools, sources, reproducibility detail, and assumptions. No findings or interpretation.
3. **Results:** objective findings, outputs, or observations, with empirical and non-empirical content clearly separated. No implications or argument.
4. **Discussion:** interpretation supported by Results, answer to the Introduction question, limitations, and relevant next steps.

## Mode Workflows

### Detector

Classify each section as `explicit`, `partial`, `inferable`, or `unavailable`. Return applicability (`high`, `medium`, or `low`), required assumptions, and one recommendation: full IMRaD, partial IMRaD, or no IMRaD. Do not turn unavailable evidence into results.

### Reviewer

Check section presence/order, boundary purity, claim traceability, empirical labels, and question→method→result→interpretation linkage. Lead with the overall verdict, then findings with severity, section, problem, impact, and specific fix. Mention strengths only when they help preserve sound structure.

### Recomposer

Extract the question and method, separate findings from interpretation, compose the four sections in exact order, expose assumptions in Methods, and validate claim traceability before handoff.

### Writer

Define the question/objective, describe the actual or explicitly analytical method, present available outputs objectively, and interpret only those outputs. If an empirical method or results are missing, do not invent them; either produce clearly labeled non-empirical work or explain why a responsible full IMRaD draft is unavailable.

Finish with the requested artifact or verdict first, followed by material assumptions, evidence gaps, and section-boundary caveats.