narumiruna/imrad-recomposer

IMRaD Recomposer

Purpose

Transform non-IMRaD material into IMRaD-structured output.

When to use

Use this skill when the user provides:

• Notes, transcripts, explanations, or fragmented analysis
• Mixed reasoning that needs separation into IMRaD sections
• Research-style analytical content requiring a production-ready IMRaD draft

When NOT to use

Do NOT use this skill when:

• The detector recommendation is "do not use IMRaD"
• The user requests only diagnosis or review
• The task is simple Q&A, tutorials, or ideation without meaningful analytical structure

Input Assumptions

• Input may be partial or noisy
• Source evidence may be incomplete
• Assumptions are allowed only when explicitly stated

Core Workflow

1. Extract the core question/problem.
2. Derive or identify method/approach.
3. Separate findings from interpretation.
4. Compose sections in exact IMRaD order.
5. Validate claim traceability and section boundaries.

Handling Incomplete Inputs

If methods or results are missing, the agent MUST:

• Infer a reasonable method only for analytical or design-oriented tasks
• State assumptions explicitly in Methods
• Use non-empirical results only when empirical results are unavailable
• Label non-empirical results as inferred, hypothetical, expected, or simulated

Truthfulness Constraints

• The agent MUST NOT fabricate empirical data, citations, experiments, measurements, or observations
• Every major claim in Results and Discussion MUST be traceable to explicit input evidence or explicit assumptions

Output Requirements

The output MUST:

• Follow IMRaD order exactly: Introduction, Methods, Results, Discussion
• Use clear section headers
• Keep boundaries clean between reporting and interpretation
• Be concise, precise, and non-redundant
• Prefer bullets where helpful

Section Specifications

Introduction

• Define context and question
• State objective or hypothesis
• MUST NOT include methods, results, or conclusions

Methods

• Describe approach, process, tools, and assumptions
• Include reproducibility details when available
• MUST NOT include interpretation or findings

Results

• Report findings objectively
• Separate empirical from non-empirical outputs
• MUST NOT include interpretation, implication, or argument

Discussion

• Interpret only what Results support
• Answer the question from Introduction
• Include limitations and concrete next steps when relevant

Failure Modes

• Mixing Results with Discussion
• Introducing claims unsupported by evidence or assumptions
• Hiding assumptions used to infer missing components
• Producing generic prose instead of structured analytical transformation