Portfolio Weekly Tracker

Problem

When you hold a portfolio of A-shares, H-shares, US stocks, B-shares, or ETFs, weekly tracking tends to become fragmented: price tables show movement but not whether each holding thesis is still valid, while announcements, earnings, operational changes, and industry events are scattered across sources and easy to miss. Portfolio Weekly Tracker treats the user's holdings, investment assumptions, and falsification thresholds as configuration, then produces an HTML weekly report built for review.

How it works

• Config-driven: Define instruments, markets, currencies, tags, and assumptions in config/holdings.json; watch_only and derivative monitoring are supported, with no default holdings.
• Hypothesis testing: Each assumption must be falsifiable by a specific number or event. States include healthy, leading, triggered, confirmed, invalid, and no_data; status is evaluated through scripts/check_thresholds.py rather than memory.
• Two-level scanning: Covers company-level announcements, filings, and operating changes, plus industry-level events affecting holdings. Price and FX anchors used for thresholds should be fetched on the report date.
• Structured output: The report follows a fixed order, including core conclusions, threshold overview, market snapshot, company-level updates, industry scan, multi/short ranking, upcoming event calendar, and a data gap statement, with sources and cross-checks noted.
• Data discipline: Prioritizes announcements, official channels, and regulated data providers; recommends dual-source verification; and explicitly degrades when APIs are unavailable instead of filling missing values with stale or estimated data.

Boundaries and cautions

This skill is useful for regularly reviewing a self-managed portfolio, checking assumptions, and organizing company information, but it is not investment advice and a triggered status should not be treated as an automatic sell order. Missing data should be marked as pending verification, not as no important events; holdings and assumptions are private user configuration and should not be hardcoded or emitted as defaults.