Summary

This tool reverse-maps investment themes' supply chains to identify overlooked upstream bottleneck stocks, distilled from public methodology. Targets investors seeking an edge by surfacing early-stage candidates, avoiding crowded names.

How It Works

Analyzes themes via 5-layer reverse supply chain breakdown, applying 9 bottleneck archetypes to pinpoint upstream companies. Generates self-contained HTML reports with supply chain visualizations, candidate leaderboards (multi-axis valuation/momentum), and actionable insights. Core differentiator: engineering discipline with multi-layered defenses against LLM failures (hallucination, bias) for data trustworthiness and judgment accountability.

Quick Start & Requirements

• Installation: Clone repo to Codex/Claude skills dir (e.g., ~/.codex/skills/serenity-bottleneck-hunter).
• Prerequisites: Requires OpenAI Codex/Claude. EODHD_API_KEY recommended (fallback yfinance for US); Web search prohibited for prices.
• Links: Repo: https://github.com/Mrjie7205/serenity-bottleneck-hunter. Key docs: SKILL.md, reference/methodology.md. Example: reference/example_commercial_space.md.

Highlighted Details

• Engineering Defenses: LLM error defenses: audits (A, A+, A++), reverse research (§A), falsification (§B), alpha feedback loop.
• Comprehensive Reporting: HTML reports: SVG supply chain graphs, leaderboards (waterfall, P/E, PEG, RS), triggers.
• Forward Validation: Emphasizes out-of-sample testing (forward_picks.csv, score_tracker.py), Alpha = stock return - theme ETF return.
• Data Integrity: Price data: scripts/price.py (EODHD/yfinance); no manual guessing. Valuation: multi-sourced (akshare, yfinance) with sanity checks.

Maintenance & Community

Independent distillation of Serenity's (@aleabittedit) methodology. No community channels or active maintainer details provided.

Licensing & Compatibility

MIT license. Independent distillation, not affiliated with methodology source.

Limitations & Caveats

Early validation phase; sample sizes may be insufficient. Valuation data is weakest; yfinance may have non-US gaps. Validation is logic-consistency/forward tracking, not audited performance.