longform-journalism

name	longform-journalism
description	Skill for transforming environmental scanning reports into longform journalism format. Use when: (1) delivering analytical reports to a broad audience beyond decision-makers, (2) restructuring complex geopolitical/technological analysis in a narrative context, (3) converting reports into Foreign Affairs/The Atlantic-level longform articles.
context	fork

Longform Journalism Transformation Skill

Overview

Transforms environmental scanning reports (integrated daily reports, topic-specific deep-dive reports) into Korean longform journalism articles. Preserves all analytical content from the source report (lossless transformation) while applying narrative structure and journalistic style to maximize readability and persuasiveness.

---

Core Principles

1. Lossless Transformation

All essential information from the source report must be included in the final article.

• All signal core facts must appear in the body text
• All sources, methodologies, and confidence data are moved to endnotes
• Inventory cross-check is mandatory to verify no omissions
• Allowed transformations: structural rearrangement, style conversion, expression refinement
• Prohibited transformations: fact deletion, meaning alteration, unsupported additions

2. Narrative First

Restructure the section-by-section enumeration of analytical reports into a narrative flow.

• Select the optimal narrative structure: chronological, causal, or contrastive
• Use Hook --> Development --> Resolution structure to engage the reader
• Integrate data and examples naturally into the narrative context

3. Endnote Separation

Separate metadata and methodology into endnotes to improve body readability.

• Body text: core facts, analysis, narrative, quotations
• Endnotes: sources, methodology, confidence levels, detailed data, some counterarguments/rebuttals
• Use [^N] markdown footnote format, numbered consecutively starting from 1

4. Style Guide Compliance

Apply all rules from final-report-style-guide.md.

• Completely remove internal codes (WF1, WF2, pSST, etc.)
• Spell out English abbreviations in full + provide Korean translation
• Convert STEEPs category codes to descriptive form
• Reference path: .claude/skills/env-scanner/references/final-report-style-guide.md

---

Transformation Workflow

Phase 1: Input Analysis

1. Read report: Read the entire source report and understand its structure
2. Determine type: Identify the report type
   • Integrated daily report: Comprehensive analysis covering multiple signals
   • Topic-specific deep-dive report: In-depth analysis on a single topic
   • Standalone workflow report: WF1 or WF2 standalone output
3. Extract inventory: Catalog all essential elements from the source
   • Signal titles (complete list)
   • Key data points (numbers, dates, ratios)
   • Sources (complete list)
   • Argumentation structure (premise-sub-premise-conclusion)
   • Counterarguments and rebuttals
   • Scenarios
   • Strategic implications

Phase 2: Narrative Construction

Design the article structure by referencing references/narrative-structure.md.

1. Generate article title: Journalistic-style title + subtitle
2. Write introduction: Apply Hook technique (800-1,200 characters)
3. Compose body sections (Parts): Cluster signals into narrative Parts
   • Apply clustering rules from references/content-mapping-rules.md
   • Assign subtitles to each Part
   • Write transition sentences between Parts
4. Write epilogue: Circular structure, open questions (600-1,000 characters)
5. Compose endnotes: Order: sources -> methodology -> confidence -> data

Phase 3: Style Refinement

1. Apply style guide
   • Read .claude/skills/env-scanner/references/final-report-style-guide.md
   • Convert internal codes (WF1 -> general environmental scanning, pSST -> confidence, etc.)
   • Convert English abbreviations to full form
2. Adjust tone
   • Foreign Affairs/The Atlantic style: authoritative yet accessible prose
   • Remove unnecessary academic jargon; explain essential academic concepts in prose
   • Convert tables -> prose or narrative enumeration (core data tables may be retained)
3. Refine sentences
   • Minimize passive voice
   • Prioritize specific figures/examples
   • Maintain paragraph length of 3-5 sentences
   • Provide explanation on first occurrence of technical terms

Phase 4: Lossless Verification + Output

1. Inventory cross-check
   • Verify all items from the Phase 1 inventory exist in the body or endnotes
   • If missing items are found, return to Phase 2 to supplement
2. Internal code verification
   • Search final file for "WF1", "WF2", "pSST", "SKEL-", "SIG-" -> confirm 0 matches
3. Endnote integrity verification
   • Verify [^N] numbering is consecutive
   • Verify 1:1 match between body references and endnote definitions
4. Save file
   • Filename: {source filename with type changed to longform journalism}.md
   • Save path: same directory as the source
5. Open for final review
   • Open file with open command and request user confirmation

---

Quality Checklist

A. Structure

• Is the article title in journalistic style (different from the report title)?
• Does the introduction have a Hook (scene/statistic/metaphor)?
• Is the body divided into Parts?
• Does the epilogue echo the introduction (circular structure)?
• Are endnotes separated after the body text?

B. Lossless

• Do all signal titles from the source appear in the body or endnotes?
• Are key data points (numbers, ratios, dates) preserved?
• Are all sources recorded in the endnotes?
• Is the argumentation structure (premise-sub-premise-conclusion) expressed narratively?
• Are counterarguments and rebuttals included?

C. Style

• Are there 0 instances of internal codes (WF1, WF2, pSST)?
• Do English abbreviations have Korean translations alongside?
• Is the tone at Foreign Affairs/The Atlantic level?
• Is passive voice minimized?
• Are paragraphs 3-5 sentences long?

D. Endnotes

• Are [^N] numbers consecutive starting from 1?
• Do body references and endnote definitions match 1:1?
• Is endnote order: sources -> methodology -> confidence -> data?
• Are endnotes free of internal codes?

E. Readability

• Are technical terms explained on first occurrence?
• Are transitions between Parts natural?
• Is the introduction within 800-1,200 characters?
• Is the epilogue within 600-1,000 characters?

F. Final Verification

• Does the filename follow the naming convention?
• Is the save path the same as the source?
• Has the file been delivered to the user via open command?

---

External References

This skill reads the following files by reference (does not copy):

Reference File	Purpose
.claude/skills/env-scanner/references/final-report-style-guide.md	Internal code conversion, English abbreviation rules

---

Input/Output Example

Input

env-scanning/integrated/reports/daily/중국_군부숙청_분석보고서_2026-02-04.md

Output

env-scanning/integrated/reports/daily/중국_군부숙청_장편저널리즘_2026-02-04.md

---

Limitations

• Cannot modify or delete facts from the source report
• Cannot add information not present in the source (narrative connecting sentences are allowed)
• Cannot leave internal workflow codes in the final output
• Cannot skip endnote numbers (must be consecutive: 1, 2, 3, ...)

---

Version

• Skill Version: 1.0.0
• Compatible with: Claude Code v1.0+
• Last Updated: 2026-02-04