Sector Radar Scan

The Problem

In A-share sector trading, the hard part is not confirming a move after it happens, but identifying which catalysts are forming, which price signals are already confirmed, and which funding clues are still in early accumulation before a sector enters a momentum phase. Sector Radar Scan is designed for left-side market reconnaissance, turning fragmented data from Tonghuashun, SunSirs/BCYF, financial news, and research reports into a repeatable pre-market scan.

How It Works

The skill uses a WebFetch structured-page + WebSearch unstructured-info architecture with built-in fallbacks. The core workflow includes:

• Exclude already-hot sectors: It first checks industry and concept sector performance, moving sectors that meet rise thresholds or limit-up concentration rules into an avoidance list.
• Scan catalysts: It searches for export controls, raw-material price spikes, capacity exits, strategic mineral controls, exhibitions, policy implementation, product launches, and mass production, then looks for chain signals such as “policy control → supply contraction → price spike → domestic substitution.”
• Verify prices and funding: For price-sensitive commodities, it searches for latest price and momentum data. Funding signals are cross-checked across institutional research, northbound flows, ETF share changes, and bottom-volume patterns.
• Score six dimensions: It outputs a total score based on position safety, catalyst certainty, price momentum, funding accumulation, timing urgency, and industry-chain fit, while marking single-source, multi-source, or conflicting evidence.
• Log and revisit: Each scan is saved to ./sector-radar/{YYYYMMDD}.md, and later runs update historical signal states to avoid re-recommending sectors that have already exploded or lost their logic.

Boundaries

This is better suited for pre-market scanning on trading days than for intraday chasing or real-time trade execution. If both WebFetch and WebSearch fail, the skill enters emergency mode, does not fabricate data, and only reviews the last valid scan; it does not score when no cache exists. The output is fact-based signal synthesis, not buy or sell advice.