Formal academic output profiles

MedPaper selects an output profile before concept validation or drafting. The domain registry is the canonical source for names, section order, typical length, concept requirements, and writing prerequisites. The constraint engine adds machine-enforced structure, evidence, voice, overlap, reference, and word-count boundaries.

Supported profiles

Key	Intended output	Concept contract	Typical words	Statistical hooks
original-research	Clinical or empirical journal article	novelty + selling points	3,000	yes
systematic-review	Reproducible qualitative evidence synthesis	novelty + selling points	4,000	yes
meta-analysis	Quantitative evidence synthesis	novelty + selling points	4,500	yes
case-report	Single case or case series	novelty + selling points	1,500	no
review-article	Narrative or invited review	novelty + selling points	5,000	no
letter	Correspondence or brief commentary	novelty	500	no
research-proposal	Grant, protocol, or institutional plan	background + gap + objectives + methods + expected impact	5,000	no
project-closeout-report	Final delivery and variance report	baseline + methods + deliverables	5,000	no
student-paper	Course paper, capstone, or assessed essay	background + question/thesis	3,000	no
conference-paper	Full proceedings manuscript	novelty + selling points	3,500	yes
thesis-dissertation	Degree thesis or dissertation	background + gap + question + methodology	40,000	yes
arxiv-preprint	Versioned repository preprint	novelty + selling points	6,000	yes
other	Explicitly customized formal output	novelty	2,000	no

The statistical-hook column controls B8, B11, and B16. Common provenance, anti-AI, language, citation, paragraph, and cross-reference checks continue to run for every profile. A skipped type-specific hook remains visible in the audit record as not applicable.

Profile-specific integrity rules

Research proposal

• Keep proposed activities and expected impact distinct from completed work or observed results.
• Make objectives, design, milestones, ethics, data lifecycle, resources, and risks internally traceable.
• State assumptions behind budget and feasibility claims.

Project closeout report

• Preserve the original baseline and success criteria.
• Separate delivered artifacts, measured outcomes, deviations, corrective actions, remaining risks, and handoff obligations.
• Link status and financial claims to dated records; do not silently rewrite the approved scope after observing the outcome.

Student paper

• The student remains the author and verifies all sources, quotations, calculations, and conclusions.
• Record course-rubric alignment and the institution's permitted AI-use rule.
• Never generate a false reading record or claim that the student personally reviewed a source they did not inspect.

Conference paper

• Enforce the venue's page, anonymity, artifact, and presentation constraints.
• Keep the paper, abstract, poster, slides, and demo consistent while avoiding duplicated or unsupported claims.

Thesis or dissertation

• Treat chapters as dependent artifacts with a single question, method, data, terminology, and contribution contract.
• Preserve committee decisions, ethics approvals, scope changes, and data provenance across revisions.

arXiv or repository preprint

• State the review status without implying peer-review acceptance.
• Record data/code availability, licenses, access restrictions, persistent identifiers, reproduction instructions, and material version changes.

Creating a project

Use the same consolidated project facade for all profiles:

project_action(
  action="create",
  name="Prospective feasibility protocol",
  paper_type="research-proposal",
  workflow_mode="manuscript"
)

get_paper_types returns the registry exposed by the domain layer. The MCP UI enum is generated from the same registry, and cross-layer tests prevent the legacy compatibility constant from drifting.

Recording exemplar use

An exemplar may calibrate organization or reporting but cannot supply evidence or citation credit:

project_action(
  action="exemplar_usage",
  exemplar_source_id="PMID:12345678",
  exemplar_roles_json='["structure", "methodology"]',
  exemplar_target_sections_json='["Methods", "Discussion"]',
  exemplar_purpose="Compare reporting order; verify every claim independently."
)

The resulting .audit/exemplar-usage.yaml fixes evidence_eligible, citation_credit, and verbatim_copy to false. Code rejects prohibited roles and refuses new records if the policy artifact has been tampered with.