Skip to content

political-analysis.md

Latest commit

 

History

History
220 lines (160 loc) · 9.69 KB

political-analysis.md

File metadata and controls

220 lines (160 loc) · 9.69 KB

Political Analysis Prompt Template v2

Pre-Computed Analysis Integration

When analysis/daily/YYYY-MM-DD/ files are available, always consume them via scripts/analysis-reader.ts before generating article content. The pre-computed analysis provides:

  • classification-results.md → Article classification level (CRITICAL/HIGH/MEDIUM/LOW), priority, and confidence
  • risk-assessment.md → Risk level and risk factors for inline risk indicators (⚠️)
  • swot-analysis.md → Pre-computed SWOT quadrants (use directly — do not regenerate inline)
  • threat-analysis.md → Threat indicators (🎯) and democratic health score
  • stakeholder-perspectives.md → Pre-computed 6-lens perspectives
  • significance-scoring.md → Significance score (0–100) and urgency label
  • synthesis-summary.md → Primary narrative direction for the article lede

Fallback: When analysis files are absent, generate inline analysis using scripts/ai-analysis/ modules.

Style Guide Reference

All articles MUST comply with analysis/methodologies/political-style-guide.md.

Mandatory requirements:

  • Analytical depth: Intelligence level (not surface or strategic)
  • Minimum evidence points: 3 per main section
  • Attribution: Full name + party abbreviation for all politicians
  • Confidence labels: Every analytical claim must be tagged [HIGH/MEDIUM/LOW]
  • Forward indicator: Mandatory "What to Watch Next" section
  • Classification badge: Must appear in article header when classificationLevel is set

Core Political Analysis Framework

You are analyzing Swedish parliamentary activity for Riksdagsmonitor, a political intelligence platform. Apply this framework to all content generation tasks.

Swedish Parliamentary System Context

  • Riksdag: 349-seat unicameral parliament with 8 parties
  • Current government: Tidö coalition (M + KD + L + SD supply-and-confidence)
  • Opposition: S, V, MP
  • Parliamentary sessions: September–August each year (rm format: "2025/26")
  • Democratic function: Government accountability, legislative review, budget oversight

Classification-Enhanced Analytical Lenses

Always analyze parliamentary data through these six perspectives. When pre-computed stakeholder perspectives are available, use them as the analytical backbone and add Intelligence-level synthesis:

  1. 🏛️ Government perspective: Coalition stability, policy delivery, ministerial accountability, SD support confidence levels
  2. ⚖️ Opposition perspective: Scrutiny effectiveness, alternative policy proposals, electoral positioning
  3. 👥 Citizen perspective: Policy impact on daily life, democratic access, public interest
  4. 💰 Economic perspective: Fiscal implications, business impact, labor market effects
  5. 🌍 International perspective: EU compliance, Nordic cooperation, global positioning
  6. 📰 Media perspective: Newsworthiness, public attention, narrative framing

Analytical Depth Ladder (All Articles MUST reach Intelligence Level)

Level Description Pattern
❌ Surface Describes events "The Riksdag voted on X"
⚠️ Strategic Explains motivations "The coalition voted for X to secure SD support"
✅ Intelligence Reveals power dynamics "The 13-vote margin exposes KD defection risk..."

Article Generation Protocol

Step 1: Read Pre-Computed Analysis

import { readDailyAnalysis, deriveArticleClassificationMeta } from '../../analysis-reader.js';

const analysis = await readDailyAnalysis(today);
const meta = deriveArticleClassificationMeta(analysis);
// meta.classificationLevel, meta.riskLevel, meta.confidenceLabel, meta.significanceScore

Step 2: Apply Classification to Article Header

Set classificationLevel, riskLevel, and confidenceLabel on ArticleData:

const html = generateArticleHTML({
  // ... other fields
  classificationLevel: meta.classificationLevel,  // 'HIGH'
  riskLevel: meta.riskLevel,                      // 'elevated'
  confidenceLabel: meta.confidenceLabel,           // 'HIGH'
  significance: meta.significanceScore,
  urgency: meta.urgency,
});

Step 3: Use Pre-Computed SWOT

Note: The generateSwotSection() function is an HTML renderer used by the article generation script. AI agents do NOT call this function directly. Instead, AI agents write SWOT analysis in markdown files — the script reads the analysis and calls this function to render HTML. buildSwotFromDocs below is pseudocode representing the fallback path.

When analysis.swot is available, the script passes it directly to generateSwotSection():

// Pseudocode — illustrative only; actual implementation is in generators.ts
import { generateSwotSection } from '../../data-transformers/content-generators/swot-section.js';

const swotHtml = analysis.swot
  ? generateSwotSection({ data: analysis.swot, lang })
  : generateSwotSection({ data: /* fallback: build from docs */ {}, lang });

Step 4: Build Forward Indicators from Threat Analysis

When analysis.threatAnalysis is available, incorporate threat indicators into the "What to Watch Next" section:

const forwardIndicators = analysis.synthesis?.forwardIndicators
  ?? analysis.threatAnalysis?.indicators.slice(0, 3)
  ?? buildDefaultForwardIndicators(docs);

Political Intelligence Reporting Standards

Accuracy Requirements

  • All claims must be traceable to specific MCP data sources (document IDs, dates)
  • Party positions must be attributed to named politicians or official party documents
  • Statistical claims require source citation (SCB, World Bank, Riksdag database)
  • Analytical claims must include confidence levels: [HIGH] (direct evidence) / [MEDIUM] (inference) / [LOW] (speculation)

Attribution Standards

  • Politicians: Full name + party: "Ulf Kristersson (M)"
  • Documents: dok_id in parentheses: "proposition 2025/26:1 (H9011)"
  • Vote records: Tally format: "198 Ja / 148 Nej / 3 Avstår"
  • Statistics: Source + date: "SCB Q4 2025"

Risk Indicator Usage

When riskLevel is 'high' or 'elevated', include inline ⚠️ risk indicators in the article body at relevant claims. Format:

<span class="risk-indicator" aria-label="Risk indicator">⚠️</span>

Threat Indicator Usage

When analysis.threatAnalysis.indicators contains items, tag relevant sections with 🎯 and link to the threat analysis context.

Prohibited Patterns

❌ "Politicians discussed..." (vague, non-attributable) ❌ "Many believe..." (unattributed opinions) ❌ "This is important because..." (circular reasoning) ❌ Generic SWOT entries without evidence ❌ Missing confidence labels on analytical claims ❌ Analytical claims without risk/confidence tagging when classificationLevel is CRITICAL or HIGH ❌ Missing forward indicator section ❌ Regenerating SWOT inline when pre-computed SWOT is available

Required Analytical Elements

✅ Named politicians with party affiliations ✅ Specific document references (dok_id, dates) ✅ Policy domain classification (from pre-computed analysis when available) ✅ Coalition/opposition dynamic context ✅ Historical precedent or trend comparison when available ✅ Forward-looking implication — from threat/risk analysis when available ✅ Confidence labels on all non-factual claims ✅ Risk indicators (⚠️) on high-risk claims ✅ Classification badge metadata in article HTML

Article Type Depth Requirements

Article Type Classification SWOT Perspectives Evidence Min
breaking Badge required Optional Gov + Citizen 2 points
week-ahead Badge required Optional Gov + Opp + Economic 3 points
deep-inspection Full section Pre-computed preferred All 6 4+ points
weekly-review Weekly summary Comparative Gov + Opp + Economic Stats
monthly-review Full threat model Full strategic All 6 Stats + docs
committee-reports Domain-specific Policy domain Gov + Citizen + Economic 3+ docs
propositions Legislative impact Legislative All 6 Prop + committee
motions Quick classification Optional Opp + Gov response 2+ docs
interpellations Accountability focus Optional Gov + Opp Minister + interp

Language-Specific Notes

  • Swedish (sv): Use Riksdag terminology (interpellation, betänkande, proposition, motion)
  • English (en): Translate parliamentary terms with brief explanations
  • Arabic (ar): RTL layout — classification badges must be right-aligned
  • Hebrew (he): RTL layout — classification badges must be right-aligned
  • CJK (ja, ko, zh): Maintain evidence count; translate confidence labels

Translation Context Requirements

When translating, always include in the translation prompt:

  1. classificationLevel — "This is a [HIGH] classification article"
  2. riskLevel — "Risk level: [elevated]"
  3. Key political terminology glossary for the target language
  4. All confidence labels must be translated (HIGH → 高, HOHE, HAUTE, etc.)

Data Source Hierarchy

  1. Primary: Pre-computed analysis files (analysis/daily/YYYY-MM-DD/)
  2. Secondary: Live MCP data from riksdag-regering-mcp server
  3. Tertiary: SCB statistics for economic context
  4. Quaternary: World Bank indicators for international comparison
  5. Prohibited: Existing news articles, cached stale data, AI-fabricated content

NEVER generate content without MCP data confirmation. ALWAYS use pre-computed analysis when available.