thematic-scout

name: thematic-scout description: Research engine that runs the full briefing-pipeline analysis stack (yfinance technicals + E*TRADE chains + third-party recommendations + earnings calendar) across a curated set of thematic universes (semis/AI, memory, nuclear, powergrid/infra, optical AI, quantum, cybersecurity, plus user-defined). For each ticker, produces a BUY / WATCH / AVOID verdict with entry strategy (shares vs CSP) and concrete order ticket. Output is a thematic research report — separate from the daily briefing's "what to do today on existing positions" focus, this skill answers "what should I be aware of in these spaces." version: 1.0

Thematic Scout

The daily briefing manages what you already own. This skill goes outbound — researching candidate names across themes you've flagged as interesting (semis/AI, nuclear energy, memory, powergrid/infra, optical AI, quantum, cybersecurity). For each candidate it runs the same analysis stack we use for held positions: yfinance for RSI/IV-rank/200-SMA/drawdown, the etrade-chain-fetcher for real CSP entry prices, third-party recommendations for rec tier, and the earnings calendar for binary-event guards.

Inputs

# references/theme_universes.yaml — extend by editing this file
themes:
  semis_ai:
    name: "Semiconductors & AI"
    anchors: [NVDA, AMD, AVGO, TSM, ARM, ASML]
    etfs: [SMH, SOXX]
    notes: "Anchor: Claude, GPT, hyperscaler capex cycle"
  memory:
    name: "Memory"
    anchors: [MU, WDC, SNDK, STX]
  nuclear:
    name: "Nuclear Energy"
    anchors: [CCJ, BWXT, NRG, VST, CEG, OKLO, SMR, NNE]
  powergrid_infra:
    name: "Powergrid & Infrastructure"
    anchors: [VRT, GEV, ETN, EMR, ABB, HUBB, PWR]
  optical_ai:
    name: "Optical AI / Networking"
    anchors: [COHR, LITE, FN, AAOI, NVDA]
  quantum:
    name: "Quantum Computing"
    anchors: [IONQ, RGTI, QBTS, IBM]
  cybersecurity:
    name: "Cybersecurity"
    anchors: [CRWD, PANW, ZS, S, NET, OKTA, FTNT]

Per-ticker analysis (same stack as daily briefing)

For each anchor ticker the scout pulls:

Verdict per ticker

Same decision tree as long-term-opportunity-advisor's ADD logic, extended:

For BUY and CSP ENTRY verdicts, output includes a concrete order ticket using the etrade-chain-fetcher for real chain prices — same rule as the daily briefing: never yfinance for tradeable chain data.

Output

A markdown report grouped by theme:

# Thematic Scout Report — 2026-05-11

## 🔭 Semiconductors & AI

### ✅ BUY · NVDA · $215.20
- RSI 52 (neutral), IV rank 41, drawdown 4% from 52w high
- 3rd-party: BUY (tier 4, 1d fresh)
- Earnings: 2026-08-26 (108d away)
- **Strategy:** Already at decent levels; consider CSP at $195 strike (10% OTM)
  via 35-DTE Fri Jun 12. Live mid $4.20 (~6% annualized).
- **Why:** ...

### 👀 WATCH · AMD · $463.65
...

## 🔭 Memory
### ✅ BUY · MU · $797.88
...

When to run

This skill is research, not daily trading. Run it:

• Weekly (e.g., Saturday morning) to refresh the watchlist
• Whenever you want to widen the scope beyond held positions
• After a major catalyst in a theme (earnings season, FOMC, geopolitical event)

Scheduled via Cowork as thematic-scout-weekly. Output lands at ~/Documents/briefings/scout_YYYY-MM-DD.md.

What it does NOT do

• Place orders. Like the daily briefing, it's read-only.
• Replace your own thesis. The verdicts are mechanical — they say "the math is OK / not OK", not "this is a good business."
• Predict direction. RSI extremes and rec changes are conditions, not forecasts.

Why a separate skill

• Different cadence than the daily briefing (weekly research vs every-morning portfolio management)
• Different universe (outbound research vs held positions)
• Different output (thematic report vs portfolio action list)
• Same analysis primitives though — reuses etrade-chain-fetcher, the technicals fetcher pattern from snapshot_inputs.py, and the verdict logic from long-term-opportunity-advisor

Hard rules

1. Chain data for any concrete CSP entry MUST come from etrade-chain-fetcher. No yfinance for tradeable prices.
2. Fail closed. If E*TRADE is unreachable, the report renders without entry tickets and flags "broker unreachable — verify before placing."
3. Earnings guard. No CSP entry tickets when next earnings is inside the contract's life.
4. Concentration awareness (optional, when run against the live portfolio): if the user already holds 10%+ NLV in a name, the verdict downgrades to WATCH (don't add).