AI Agent Hub
Back to skills
Five-Layer Sieve Investment Analysis icon

Five-Layer Sieve Investment Analysis

Professional Updated 2026.08.29

Paste the following prompt into your AI chat to install this skill:

Please follow https://skillhub.cn/install/skillhub.md to install @user_a23cd6e1/five-layer-sieve into your AI assistant.

About this skill

Problem Addressed

  • Single-stock research often mixes data, fundamentals, valuation, and decisions, which can reward first-impression bias and silently force contradictory findings into one conclusion.
  • For engineers, the core issue is not only accuracy but auditability: which inputs came from scripts, which came from search, and where one layer contradicts an earlier one.

How It Works

  • Multi-head attention style orchestration: maps the ticker into Query, layer rules into Key, and layer conclusions into Value; Head 0 through Head 5 cover data validation, positioning, drivers, conflicts, valuation, and decisions.
  • Script-first evidence: runs existing scripts before falling back to web_search, reducing the chance that second-hand descriptions are treated as primary data.
  • Cross-layer consistency checks: after each Head, the skill checks whether its conclusion conflicts with earlier outputs; if conflict exists, it pauses for user judgment rather than auto-reconciling.
  • Higher-order synthesis: full reports continue into Head 6 to produce a more advanced forecast and a structured Markdown output.

Boundaries

  • Best suited for investment-analysis orchestration on tickers with clear financial data and industry context, not for trade execution or real-time quotes.
  • Real-time pricing, compliance opinions, and transaction decisions still need external systems and human review.

Use Cases

  • Before a stock review, enter a ticker and intent to generate layered data, fundamental, valuation, and decision analysis.
  • When only financial and industry data are needed, run validation and basic layers to pull first-hand data without valuation.
  • For disputed valuations, read valuation output and check whether model conflicts trace back to parameter assumptions.
  • When assessing management quality, run only the management sublayer to output catalysts and confidence labels.

Best For

  • Equity researchers who need to split financials, industry comparison, valuation, and decisions into inspectable layers before stock reviews.
  • Quant strategy authors who want cross-layer consistency checks and explicit model conflicts instead of silent reconciliation.
  • Financial analysts who need to query financial and industry data without automatically moving into valuation or trading advice.
  • Individual investors who want a complete, sectioned, and reviewable investment-analysis framework around a stock ticker.