Skip to content
mif-docs

Skill reference: academic

Skill reference: academic

Section titled “Skill reference: academic”

The academic skill authors one document genre: a formal academic research report — a scholarly write-up for a research or technical-expert audience that will scrutinize method, evidence, and the limits of each claim before accepting it. This reference describes what that document type is, how the skill produces one, when it earns its place, and the provenance behind it.

PropertyValue
AuthorsA formal academic research report
Purpose groupScholarly writing
MIF conceptTypesemantic
Target MIF level3
Primary sourceICMJE Recommendations (numbered/Vancouver mode); APA 7th Edition (author-date mode)

What this document type is

Section titled “What this document type is”

An academic report follows the general IMRaD convention — Introduction/ Background, Methods, Results (here, Findings), and Discussion — used broadly across scientific and technical scholarship. Its defining trait is mandatory citation discipline applied in one of two selectable modes: author-date (APA 7th Edition) or numbered (Vancouver/ICMJE), picked once and applied consistently for the whole document. Every claim in Findings must trace to a cited, verifiable source; an uncited claim is a defect, not a stylistic choice. The report also carries an Abstract, states its Method (how evidence was gathered and verified, with optional APA sub-sections for Participants, Materials, Procedure, and Analysis), and closes with Discussion — including honestly stated limitations — and a full References section.

This is distinct from an informal, prose-driven alignment narrative (a google-design-doc), from a practitioner report built around a mandatory options-vs-criteria comparison table (an engineering report), and from a single already-made, immutable decision record (an adr). It reasons from evidence to interpretation rather than recording a decision or proposing one, so it projects to MIF as semantic content at Level 3.

How the skill produces one

Section titled “How the skill produces one”

academic is a genre skill: it carries the IMRaD pattern and citation discipline as durable instructions plus exemplars, and writes the artifact over a MIF floor so the result is at once a human-readable scholarly report and a machine-conformant unit.

  • Pattern, made operational. The skill encodes the six-section IMRaD shape, the selectable citation mode (author-date vs. numbered), the required tables/figures-as-Mermaid convention, and the rule that hedges contested evidence and states limitations rather than laundering them. Fabricating a citation, or leaving a claim uncited, breaks conformance regardless of citation mode chosen.
  • Exemplars set the bar. Like every genre in the suite it ships good-l1.md (the MIF Level-1 floor), good.md (the Level-3 target), bad.md (a counter-example that states findings as fact with no citations), and evals/evals.json. The check-exemplars gate proves good-l1.md validates at L1 and good.md at Level 3.
  • MIF projection. The document is authored with MIF frontmatter (via the shared mif-frontmatter substrate) and a conceptType of semantic, reflecting that the doc is a claim set with method and evidence rather than a time-bound event or step sequence. mif-validate proves the Markdown ↔ JSON-LD round-trip is lossless before the document is considered done.

When it is beneficial

Section titled “When it is beneficial”

Reach for academic when the deliverable is a scholarly write-up for a research or technical-expert audience that demands traceable evidence and explicit method — a literature synthesis, an evaluation study, or any report whose claims must survive peer scrutiny of both the evidence and how it was gathered. It rewards teams that need every finding attributable and every disputed result flagged rather than smoothed over.

Do not use it for an informal, pre-build trade-off narrative — that is google-design-doc: conversational, alternatives reasoned in prose, no IMRaD structure or citation-style requirement. Do not use it for a practitioner decision report built around a mandatory options-vs-criteria comparison table — that is engineering: its center of gravity is the Trade-offs table, not scholarly method and cited evidence. Do not use it for a single already-made, immutable decision record — that is an adr.

Example

Section titled “Example”

A report titled “Retrieval Strategy and Evaluation in Retrieval-Augmented Generation: A Synthesis Report” opens with a 150-250 word Abstract stating the question, method, and principal findings, then states in Background that closed-book language models cannot incorporate long-tail or newly changed facts. Method explains that evidence came from three primary-source papers verified by reading the originals rather than downstream citations. Findings then walks three themes each grounded in a real, cited publication — Lewis et al. (2020) on the RAG architecture itself, Karpukhin et al. (2020) on dense passage retrieval outperforming sparse lexical retrieval, and Es et al. (2023) on RAGAS’s reference-free evaluation of faithfulness and relevance. Discussion ties the three findings together (retrieval quality bounds what the generator can faithfully ground, and evaluation determines whether that faithfulness is actually measured) and states the report’s limitation honestly: it synthesizes published papers rather than running a new empirical comparison. References lists the three cited papers in full, author-date style.

Provenance & citations

Section titled “Provenance & citations”