sector-research-skill

name	sector-research-skill
description	Use this skill when the user asks to analyze an investment sector/theme, build a public-company universe, rank candidates, and optionally create/update multiple Notion Assets pages using the structured asset template.

Sector Research Skill

Core Rules

• Do not fabricate companies, tickers, suppliers, customers, valuation metrics, market sizes, or thesis claims.
• Prefer primary sources and official filings/IR material; label source quality and uncertainty.
• Search Notion Assets before deciding create vs update for every company.
• Ask confirmation before creating/updating multiple Notion pages or making bulk changes.
• For existing assets, default to appending a dated sector update; never silently overwrite existing thesis/memo content.
• Keep domains narrow and specific; do not tag every adjacent theme.
• If primary sources are unavailable for an asset, either skip it or mark confidence Low.

When To Use

Use when the user asks for:

• Sector/theme research, e.g. AI optical transceivers, Japan robotics, nuclear supply chain.
• Creating a watchlist from a sector.
• Comparing many companies in an ecosystem.
• Supplier/customer maps across an industry.
• Ranking sector candidates and creating multiple Notion Assets pages.

For one company only, use asset-research-skill instead.

Required Clarification

Ask only blocking questions. Clarify if missing/ambiguous:

• Sector/theme and boundaries.
• Geography/market scope.
• Public companies only vs private ecosystem names allowed.
• Maximum number of companies/pages to create.
• Intent: watchlist, ranking, supplier map, valuation screen, earnings review, or deep-dive batch.
• Whether to create Notion pages after shortlist approval.

If the user gives a clear theme and no limit, assume public listed companies, global scope where relevant, and start with a shortlist of 5–10 candidates before asking for write confirmation.

Workflow

1. Scope The Sector

Define:

• Sector thesis / demand driver.
• Value chain boundaries.
• Included/excluded subsegments.
• Geography, currency, and market focus.
• Known anchor customers/suppliers/regulators.

2. Build Initial Universe

Use source hierarchy:

Tier 1 — Primary:

• Company annual reports / 10-K / 20-F / securities reports.
• Quarterly results, investor presentations, transcripts.
• Official company websites and IR pages.
• Regulatory filings.

Tier 2 — Industry/ecosystem:

• Competitor filings.
• Supplier/customer disclosures.
• Government/regulatory docs.
• Credible industry reports.

Tier 3 — Secondary:

• News articles, analyst summaries, specialist blogs, market-data sites.

For each candidate capture:

• Company name, ticker, exchange, country, currency.
• Role in value chain.
• Why it is relevant to the theme.
• Main customers/suppliers/partners when source-backed.
• Market cap and revenue scale with source/date.
• Source confidence: Low / Medium / High.

3. Search Notion Assets First

For every shortlisted candidate:

1. Search Assets / Investments DB by company name and ticker.
2. If clear match: retrieve properties; read page body only if needed for update/comparison.
3. If multiple matches: stop and ask user to choose.
4. If no match: mark as create candidate.

Output a create/update/skip table before any writes.

4. Sector Map

Produce a concise sector overview:

• Value chain map.
• Demand drivers and cyclicality.
• Competitive structure and concentration.
• Customer/supplier power.
• Regulation/geopolitics.
• Technology shifts.
• Key risks and red flags.
• Source log.

5. Rank Candidates

Rank with an evidence-first scoring grid:

• Theme relevance: 0–20.
• Business quality / moat: 0–15.
• Financial quality: 0–15.
• Growth/catalysts: 0–15.
• Valuation / margin of safety: 0–15.
• Balance sheet / dilution risk: 0–10.
• Source confidence: 0–10.

Final score: 0–100 with confidence Low / Medium / High.

6. Parallel Asset Research

For 3+ assets, use subagents when helpful to avoid context bloat:

• Launch one subagent per asset or per small group.
• Instruct each subagent to return only structured facts, sources, bull/bear, valuation assumptions, exit triggers, and uncertainties.
• Main agent remains responsible for final synthesis, Notion writes, and avoiding duplicates.
• Do not let subagents write to Notion directly unless explicitly instructed and safe.

Subagent return format:

## Asset
- Name / ticker / exchange / country / currency
- Business role in sector
- Key facts with sources
- Customers / suppliers / ecosystem with source quality
- Financial snapshot
- Valuation snapshot
- Bull case
- Bear case
- Buy modelisation assumptions
- Exit thesis / news to watch
- Proposed Notion properties
- Source log
- Confidence / unknowns

7. Propose Notion Changes

Before bulk writes, show:

## Proposed Notion changes

Sector/theme:

Pages to create:
- Company — ticker — proposed Status/Market/Market Cap/Domain/Score

Pages to update:
- Company — ticker — properties to update — content action

Pages skipped / low confidence:
- Company — reason

Content format:
- Structured asset template with real columns and real Notion tables

Confirm?

Stop for confirmation before creating/updating multiple pages.

8. Write To Notion

For each confirmed asset:

• Use asset-research-skill Notion Assets Page Layout.
• Create real Notion column_list, column, table, and table_row blocks where supported.
• Include a real supplier/customer/ecosystem table block; do not use a dense paragraph or plain markdown pipe table unless the API genuinely cannot create table blocks.
• Include a numeric Buy Modelisation / Market-cap scenarios table block with starting-point, bear/base/bull, valuation, and action/trigger rows. Each scenario must flow logically from revenue basis → margin/profitability → valuation multiple/method → implied market-cap range. The Implied market cap cells must contain only market-cap ranges, not prose; put caveats and buy/hold/sell logic in Action / trigger.
• If page creation rejects nested table blocks, create the page first and then append the real tables with Append block children.
• Include Exit thesis / News to watch.
• Include Sources with tier labels, URLs, and uncertainty notes.
• Set only confirmed properties.

For existing assets:

• Default action: append dated sector update.
• Do not replace full memo unless user explicitly confirms replacement.
• Do not silently overwrite properties if user did not confirm them.

Output Requirements

• Be concise but evidence-dense.
• Separate facts, assumptions, and interpretation.
• Cite source names/URLs next to important claims when possible.
• Include universe coverage limits and remaining unknowns.
• State confidence for sector overview and each asset.

Stop Conditions

Stop and ask if:

• Sector boundaries are ambiguous enough to produce the wrong universe.
• Primary sources cannot be found for core candidates.
• Notion has multiple likely matches for a company.
• User requests broad/bulk Notion writes without confirmation.
• Requested action would overwrite, archive, delete, or materially replace existing content.