heurist-finance: Skill for Claude Code

Market Health

Shared plumbing and output contract: ../_shared/analyst-core.md. The most consequential retail question isn't stock selection — it's how much equity exposure to carry right now. Most public content answers with commentary; this skill answers with a gate. The output is one of three states, borrowed from the best practice in public trading skills and worth adopting as-is:

• NEW_ENTRY_ALLOWED — regime supports adding risk; normal sizing applies.
• REDUCE_ONLY — hold what's working, no new entries, trim laggards; tighten falsifiers.
• CASH_PRIORITY — distribution/stress regime; raising cash outranks any new idea.

The state is a gate on other skills (momentum entries, screen finalists, event trades all check it), and its confidence is capped by input completeness: with inputs missing or conflicting, default to the more conservative state and say which inputs were unavailable — never average conflicting readings into false moderation.

The inputs (pull all five lanes; each can veto)

1. Index trend + distribution counting. technical_snapshot symbols=^GSPC,^IXIC (daily + weekly). Trend state vs 50/200d first. Then count distribution days — declines ≥ ~0.2% on higher volume than the prior day — over the trailing 25 sessions (compute from prices ^GSPC --period 3mo --full --out FILE; volume needs --full): 4–6 in the window is the O'Neil-school caution signal that institutional selling is underway, historically preceding intermediate corrections. Cluster timing matters more than raw count; expire days older than 25 sessions. A follow-through day (major index +1.5%+ on rising volume, day 4+ of a rally attempt) is the same school's re-entry signal after a correction — label both as practitioner methodology with real but imperfect hit rates (false FTDs happen in bear markets; the tool is exposure throttling, not prophecy).
2. Breadth (the market of stocks vs the stock market). % of members above 200d/50d MAs, new highs − new lows, advance/decline behavior; equal-weight vs cap-weight spread (RSP vs SPY via prices) is the cheap always-available proxy — a cap-weighted index at highs while the equal-weight lags badly and NH-NL is negative = narrow leadership, the classic late-stage signature. Sourcing: attempt the primary breadth statistics first via web (Barchart/StockCharts publish them); only after that attempt fails, say so and fall back to the RSP/SPY proxy + sector participation (sector-rotation), labeled as proxy.
3. Volatility regime. ^VIX level + realized vol of the index. VIX > ~28–30 with an inverted VIX term structure = stress regime (forced flows dominate); VIX grinding under ~13 for months = complacency (cheap hedges, stretched rubber band — an input to REDUCE_ONLY only in combination). The transition (VIX regime shift on rising realized) matters more than the level.
4. Credit confirmation. From macro-regime: IG/HY OAS direction (fred macro_series_history). Equity weakness without credit confirmation is usually rotation → don't gate to CASH_PRIORITY on equities alone. Credit widening + equity weakness = genuine regime risk.
5. Leadership quality. What's leading the tape (equity_screen day_gainers, sector RS from sector-rotation, news tone via yahoo news_search): offensive leadership (cyclicals, quality growth, broadening participation) vs defensive/speculative leadership (staples+utilities leading = risk-off internals; low-quality speculative froth leading = late-stage recruitment). Leadership is the tiebreaker lane when trend and breadth disagree.