# Eurhta Report > Draft a EUnetHTA Core Model HTA report chapter by chapter for an HEOR dossier: Executive Summary, Introduction, Technology Description, Health Problem, Literature Research, Clinical Effectiveness, Safety, Economic Evaluation (CEA/BIA), Ethical/Organisational/Social/Legal aspects, and Discussion. Injects PRISMA literature-review numbers and deterministic economic-model results, and cites only included studies. Use for: HTA dossier, HTA report, EUnetHTA, Core Model, health technology assessment, European HTA, joint clinical assessment, dossier chapter drafting. - Skill: `aliakseit/eurhta-report` (Agent Skill, multi-file: 15 files) - Install (CLI): `npx skillmds@latest add aliakseit/eurhta-report` - Raw SKILL.md: https://api.skillmd.com/api/skills/aliakseit/eurhta-report/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: AliakseiT (https://skillmd.com/u/aliakseit) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/aliakseit/eurhta-report --- # EUnetHTA Core Model Report Drafting Drafts a full HTA report in a dossier directory (see the plugin's `DOSSIER_FORMAT.md`), one chapter file at a time. Act as an expert HEOR researcher generating chapters of an HTA report according to EUnetHTA guidelines. ## Chapter set and file naming Files go to `chapters/-..md` (language from `dossier.yaml` `languages`, default `en`). Numbering follows the report's presentation order: | nn | Chapter | Instruction file (references/chapters/) | |----|---------|------------------------------------------| | 01 | Executive Summary | `executive-summary.md` | | 02 | Introduction and Problem Description | `introduction-and-problem-description.md` | | 03 | Technology Description | `technology-description.md` | | 04 | Health Problem and Current Use | `health-problem-and-current-use.md` | | 05 | Literature Research | `literature-research.md` | | 06 | Clinical Effectiveness | `clinical-effectiveness.md` | | 07 | Safety | `safety.md` | | 08 | Economic Evaluation | `economic-evaluation.md` | | 09 | Ethical Aspects | `ethical-aspects.md` | | 10 | Organisational Aspects | `organisational-aspects.md` | | 11 | Social Aspects | `social-aspects.md` | | 12 | Legal Aspects | `legal-aspects.md` | | 13 | Discussion and Conclusions | `discussion-and-conclusions.md` | **Drafting (dependency) order differs from numbering:** draft Literature Research (05) first, then 02, 03, 04, 06, 07, 08, 09, 10, 11, 12, 13, and the Executive Summary (01) **last** — it summarizes everything else. The user may select a subset of chapters; keep this relative order. ## Step 1 — read the dossier state - `dossier.yaml` — intervention, comparator, jurisdictions, languages, and `notes` (global context/instructions: currency, style, terminology — they apply to every chapter). - `documents/` — supporting documents (IFU, device description, protocols, reports); primary evidence alongside the literature. - `prisma/` — `pico.yaml`, `prisma-diagram.md`, `included-studies.json`, `screening.json`. If PRISMA is absent, tell the user the Literature Research and evidence chapters will be weak and offer to run the `prisma-review` skill first. - `models/inputs/*.json` and `models/runs/*.json` — economic model inputs and engine outputs (CEA, BIA, PSA, scenarios). The latest run per model is the one to report. - Existing `chapters/*.md` — human-owned; when regenerating, preserve manual edits (write alongside or show a diff, never blind-overwrite). ## Step 2 — draft each chapter For every chapter, load its instruction file from `references/chapters/` and apply these cross-cutting rules (ported from the source app's shared prompt preamble): 1. **Global context first.** Apply `dossier.yaml` `notes` to the entire chapter content, especially currency, style, and terminology. 2. **Regeneration instruction is law.** If the user asks to regenerate a chapter with a specific instruction, you MUST follow it. 3. **Structure.** Start the file with `# ` (adaptation: in the source app the title was added by the system; in a dossier each file carries its own H1). Then the body — do not repeat the title as another heading. 4. **Coherence, no repetition.** Read the already-drafted chapters and avoid redundant explanations; refer to other chapters instead (e.g., "as detailed in the Literature Research chapter"). 5. **Citation discipline.** Cite **only** studies present in `prisma/included-studies.json` (plus user-supplied documents). Mandatory references MUST be included and discussed; excluded references must NOT appear. Use the same reference numbering as the Literature Research chapter. 6. **Numbers are read, never computed.** PRISMA counts come from `prisma/prisma-diagram.md` (regenerate via the `prisma-review` skill's `scripts/prisma-counts.ts` if stale). Economic results come verbatim from `models/runs/*.json` (produced by `packages/heor-engine` via `npx tsx`). Never calculate, extrapolate, or round-trip model results yourself; if a needed run is missing, say so and stop rather than inventing values. 7. **No external knowledge as evidence.** Background framing is fine; every evidential claim traces to an included study, a supporting document, or an engine result. ## Step 3 — final review pass After all chapters are drafted: - Cross-check consistency: comparator named identically everywhere; PRISMA numbers in chapters 05/06 match `prisma-diagram.md`; ICER/BIA figures in chapters 01, 08, 13 match the latest `models/runs/` files. - List any gaps (missing evidence, assumptions used in the models) for the user. - Append to the Executive Summary and to your final message the disclaimer: *"This HTA report draft was generated with AI assistance. Expert review by qualified HEOR/regulatory professionals is required before submission."* ## Optional Claude-specific enhancement (skip on other harnesses) Independent chapters within the same dependency stage (03, 04, 09–12) may be drafted by parallel subagents, each given the dossier paths, the shared rules above, and its single chapter instruction file. Merge, then run Step 3 yourself.