heurist-finance: Skill for Claude Code

Short Thesis

Shared plumbing and output contract: ../_shared/analyst-core.md. Shorting has inverted math (losses unbounded, winners shrink as they work, position grows against you when wrong) — so a short thesis needs more than a mirrored long: it needs a reason, a catalyst, and a squeeze gate. Overvaluation alone is none of these.

The three-legged test — all three or no short

1. The reason (why it's wrong, not just rich). The durable short archetypes, roughly in order of reliability:

• Fading fundamentals dressed as growth — the forensic cluster from fundamental-forensics (run it; it's the evidence engine for this skill): revenue >30% with CFO/NI <0.5 for 2+ years, DSO rising, receivables outgrowing revenue.
• Quadratic cost structures (Chanos): the model only works if customer acquisition scales geometrically while unit economics never close — growth is the liability.
• Broken cyclical still priced as secular — the three-peaks logic: momentum peak, then price peak, then profit peak; the short window is after the price peak while consensus still extrapolates the profit line. Run the EPS sensitivity (revenue ≈ P×Q on a fixed cost base) to show what a 30% ASP giveback does to consensus.
• Structural obsolescence with a levered balance sheet — the debt turns a slow decline into an event (refinancing wall = your catalyst, check maturities via xbrl_fact_trends/filings).
• Pure valuation shorts ("it's at 40× sales") are the weakest class — they require the multiple to compress on schedule, which is a timing bet with no timer.

2. The catalyst. "Being early is being wrong" is the short seller's first law — overvalued names can double while you wait, and the borrow bill runs the whole time. Catalysts that qualify: refinancing wall inside 12–18 months, covenant proximity, lockup expiry (supply), guidance that can't be walked up, auditor/CFO departure, index deletion, regulatory decision with a date. Check filing_timeline, equity_overview (earnings dates), lockup terms via web. No catalyst inside the horizon → the honest output is "right thesis, no trade yet — here are the tripwires to watch" (and write them into the plan file as watch items).