earnings-attribution

npx skills add majiayu000/claude-skill-registry --skill earnings-attribution-faisalanjum-eventtrader --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

# Earnings Attribution Analysis **Goal**: Determine WHY a stock moved after an 8-K filing by calculating the SURPRISE (actual vs expected). **Thinking**: ALWAYS use `ultrathink` for maximum reasoning depth. **End Goal**: Build company-specific driver understanding for real-time prediction accuracy. --- ## Core Principles 1. **Comprehensiveness Over Speed**: Query ALL relevant data sources 2. **Evidence-Based Claims Only**: Every claim must cite a source 3. **Surprise-Focused**: Stock moves on actual vs expectations, not absolute results 4. **Learn from History**: Each analysis builds company-specific knowledge 5. **Self-Audit**: Validate all claims have sources before completing --- ## Confidence Levels - **High**: Multiple sources agree, clear fundamental reason, unambiguous evidence - **Medium**: Sources partially agree, conclusion fits with caveats - **Insufficient**: Sources conflict, missing critical data, no clear driver **If evidence is insufficient, say so. Never fabricate certainty.** --- ## Resources - **Output format**: [output_template.md](../../shared/earnings/output_template.md) - **Evidence audit checklist**: [evidence_audit.md](../../shared/earnings/evidence_audit.m

python3 scripts/build-thinking-index.py {accession_no}