parallax-ai-soros - SKILL.md Agent Skill

name: parallax-ai-soros description: "Applies George Soros's top-down reflexivity framework (per 'The Alchemy of Finance', 1987) to current Parallax data. Triggers: 'trade ideas based on current macro regime', 'regime-driven stock picks', 'reflexivity-lens stock ideas'. Two modes: basket mode surfaces regime themes and ranked trade ideas; single-ticker mode runs the same macro workflow and checks ticker exposure via dual channels (industry exposure AND telemetry basket theme). Third-person framing, book citation, AI-inferred from public information. NOT financial advice. NOT personalized. Accepts plain tickers or RIC format. NOT for bottom-up factor analysis (use /parallax-deep-dive). For all five profiles simultaneously use /parallax-ai-consensus."

Parallax AI Soros Profile

When not to use

Bottom-up factor scoring → use /parallax-ai-buffett
Mechanical formula screen → use /parallax-ai-greenblatt
Balance-sheet-first checks → use /parallax-ai-klarman
Cross-profile consensus → use /parallax-ai-consensus
Full macro outlook without a persona lens → use /parallax-macro-outlook
Morning brief for a portfolio → use /parallax-morning-brief

Gotchas

JIT-load _parallax/parallax-conventions.md, profile-schema.md, output-template.md, profiles/soros.md
Soros has TWO modes — basket (no ticker) and single-ticker dual-channel — sharing a macro workflow
list_macro_countries is 1 token; macro_analyst is 5 tokens per market — cap at 5 markets
get_telemetry is 1 token but distinctive — it's the second exposure channel for single-ticker mode
Single-ticker mode requires BOTH the macro workflow AND a ticker-exposure check via 2 independent channels
NEVER use first-person impersonation of Soros — always frame as "Soros-style" or "reflexivity lens identifies"
The dual-channel logic is load-bearing — single-channel flag is partial, not full match
Disclaimer verbatim per output-template.md, substituting "George Soros" for [Investor]
build_stock_universe is async (~15-40s) and may time out — fallback per conventions §4
Telemetry basket-membership resolution: ticker → basket assignment is currently best-effort via name-match against divergences[*].basket_name; per-ticker basket lookup is NOT a documented field in the get_telemetry schema. Treat Channel B (single-ticker mode) matches as best-effort pending a documented convention in parallax-conventions.md.
JIT-load _parallax/white-label/integration-pattern.md before the Pre-Render step. Loader call is load_visual_branding() (6-key visual subset; voice structurally excluded — branding["voice"] raises KeyError). Apply §5 (Branding Header) and §7 (Provenance) in Output Format.

Top-down reflexivity lens: regime identification → thematic exposure → ranked trade ideas OR single-ticker dual-channel check.

Usage

/parallax-ai-soros                         # basket mode — regime themes + ranked ideas
/parallax-ai-soros AAPL.O                  # single-ticker mode — dual-channel exposure check
/parallax-ai-soros --markets US,JP,EU      # basket mode with explicit market list

Workflow

Execute using mcp__claude_ai_Parallax__* tools.

Step 0 — JIT-load dependencies

Before the first Parallax tool call:

Load _parallax/parallax-conventions.md
Load _parallax/AI-profiles/profile-schema.md
Load _parallax/AI-profiles/output-template.md
Load _parallax/AI-profiles/profiles/soros.md

Call ToolSearch with query "+Parallax" to load the deferred MCP tool schemas.

Step 1 — Determine mode

Zero tickers → basket mode
Exactly one ticker → single-ticker mode
Multiple tickers → reject: "Soros profile takes zero or one ticker. Use /parallax-ai-consensus for multi-ticker."

Step 2 — Macro workflow (runs in BOTH modes)

Call list_macro_countries to get covered markets (1 token).
Select 3-5 tactically interesting markets. Default: US, JP, EU plus top-2 EM based on telemetry divergence. If user passed --markets, use that list.
Call macro_analyst(component=tactical) for each selected market IN PARALLEL (5 tokens each × 3-5 = 15-25 tokens).
Call get_telemetry for cross-market regime divergence signals (1 token).
Identify 1-3 regime themes where macro and telemetry agree on directional views. Examples:
- "Fed cutting + dollar weakening → EM equities and commodities"
- "Rates higher for longer → financials and defensive dividend payers"
- "Fiscal divergence US vs EU → US cyclicals overweight EU cyclicals"

Step 3 — Basket mode completion (skip if single-ticker)

IMPORTANT — thematic query scoping: build_stock_universe is async and broad queries time out. Thematic queries MUST be sector-scoped and ideally size-bounded. Good: "US energy exporters benefiting from dollar weakness". Bad: "companies benefiting from current macro regime".

For each identified theme, call build_stock_universe with a sector-scoped thematic query (5 tokens each, async). If any theme's universe times out, retry ONCE with a narrower version of the query; if it still times out, skip that theme and proceed with remaining themes.
Cap each theme's candidate list at top 20 names by composite_score from the universe response to bound token cost and avoid excess peer snapshots.
For top 3-5 names per theme (ranked by composite_score), call get_peer_snapshot (1 token each).
Apply cross-validation gate per profile-schema.md §2 Step 2 for each ticker surfaced (target_company field from peer snapshot).
Rank names within each theme by momentum + macro sensitivity.
Render as trade_ideas output (Step 5).

Step 4 — Single-ticker mode completion (skip if basket)

Call get_company_info on the input ticker (1 token) — retrieve sector, industry.
For each identified theme, call build_stock_universe with the theme's sector-scoped query (5 tokens each). Apply the same retry-once-narrower fallback as Step 3.
Dual-channel exposure check — Channel A has TWO independent sub-paths that are evaluated independently:

Channel A — Industry exposure (two sub-paths; either is sufficient to FLAG):
- Sub-path A1 — Universe membership (requires build_stock_universe): Does the ticker appear directly in any theme's build_stock_universe result? If build_stock_universe times out or returns empty, A1 is NOT_FLAGGED but Channel A is not yet resolved — continue to A2.
- Sub-path A2 — Sector/industry classification match (does NOT require build_stock_universe): Does the ticker's sector/industry classification from get_company_info match a theme's target industry or sector? (e.g., ticker is Energy sector, theme is "Energy benefiting from dollar weakness" → MATCH; ticker is Consumer Staples, theme is "Cyclicals benefiting from rate cuts" → NO MATCH.)
- Channel A result: FLAGGED if EITHER A1 OR A2 matches. NOT_FLAGGED only if BOTH are negative. A2 is evaluable independently of A1 and does not require the universe build to succeed.
Channel B — Telemetry basket theme:
- Does the ticker fall into any regime basket surfaced by get_telemetry?
- Telemetry baskets are typically named (e.g., "growth-over-value rotation," "dollar regime beneficiaries").
- If get_telemetry fails (e.g., "Admin org not configured"), Channel B is UNAVAILABLE (not NOT_FLAGGED). This distinction matters for verdict computation.
- Output: FLAGGED, NOT_FLAGGED, or UNAVAILABLE for channel B.
Cross-validation gate: after get_peer_snapshot (if called during universe resolution), cross-check target_company (the top-level field — NOT name on individual peer rows) against get_company_info's name per conventions.
Combine verdicts:
- Both channels FLAGGED → verdict match
- Exactly one channel FLAGGED, the other NOT_FLAGGED or UNAVAILABLE → verdict partial_match
- Neither flagged (both NOT_FLAGGED, or one NOT_FLAGGED and one UNAVAILABLE) → verdict no_match
- Verdict match is NEVER reached when Channel B is UNAVAILABLE — maximum single-channel verdict is partial_match

Step 5 — Render through output template

Basket mode output:

Soros-style regime themes and trade ideas

Source: Soros, G. (1987). The Alchemy of Finance. | Drobny, S. (2006). Inside the House of Money.

Reflexivity framework: identify regime break or narrative-fundamentals divergence → surface exposed industries → concentrate on high-conviction names.

Markets analyzed: <list>

## Theme 1: <name>
Macro thesis: <one paragraph — regime break or tactical opportunity>
Directional view: <long / short / rotation>
Currency/rate context: <brief>

Ranked trade ideas:
| Rank | Ticker | Industry | Momentum | Rationale |
|------|--------|----------|----------|-----------|
| 1    | XXX.N  | ...      | 7.2      | ...       |
| ...  | ...    | ...      | ...      | ...       |

## Theme 2: <name>
[same structure]

Workflow derived from: Soros (1987); Drobny (2006).
Last anchor-tested: 2026-04-07 (NVDA.O, TSLA.O) | Last legal review: pending
Tool sequence: list_macro_countries, macro_analyst × N, get_telemetry, build_stock_universe × N, get_peer_snapshot × M
Token cost: ~25-40 tokens

---
This output is an AI-inferred interpretation of George Soros's approach, derived solely from publicly available information — the cited source, Parallax factor data, and Parallax's public methodology. It is produced by the Parallax AI Investor Profiles framework. It is not financial advice, not personalized, not endorsed by George Soros or his representatives, and not a recommendation to buy or sell any security. For illustrative and educational use only. Past characterization does not guarantee future relevance. Please consult a qualified financial advisor before making investment decisions.

Single-ticker mode output:

Soros-style profile applied to <ticker>

Source: Soros, G. (1987). The Alchemy of Finance.

Reflexivity framework: top-down regime identification → thematic exposure check. Single-ticker mode answers "is this stock in an industry and basket theme the current regime analysis flags?"

Current regime themes (1-3): <list with one-line theses>

Ticker exposure check (dual-channel):

  Channel A — Industry exposure
    <ticker>'s sector/industry: <sector>
    Industry appears in theme(s): <theme names or NONE>
    Status: FLAGGED / NOT FLAGGED

  Channel B — Telemetry basket theme
    Telemetry baskets surfaced: <list>
    <ticker> falls in basket(s): <basket names or NONE>
    Status: FLAGGED / NOT FLAGGED

Profile fit: <match / partial_match / no_match>
  - match: both channels flagged (genuine thematic exposure)
  - partial_match: one channel flagged (industry OR basket, not both)
  - no_match: neither channel flagged (Soros-style lens has no current view)

Workflow derived from: Soros (1987); Drobny (2006).
Last anchor-tested: 2026-04-07 (NVDA.O, TSLA.O) | Last legal review: pending
Tool sequence: list_macro_countries, macro_analyst × N, get_company_info, get_telemetry, get_peer_snapshot
Token cost: ~25-30 tokens (single-ticker mode)

---
This output is an AI-inferred interpretation of George Soros's approach, derived solely from publicly available information — the cited source, Parallax factor data, and Parallax's public methodology. It is produced by the Parallax AI Investor Profiles framework. It is not financial advice, not personalized, not endorsed by George Soros or his representatives, and not a recommendation to buy or sell any security. For illustrative and educational use only. Past characterization does not guarantee future relevance. Please consult a qualified financial advisor before making investment decisions.

Step 6 — Emit

Output additions (white-label branding + §9.2 disclosure)

These additions apply to the rendered output ABOVE in addition to the persona-specific disclaimer shown in the output example. They are required regardless of view state.

Pre-Render — Load white-label branding

Load _parallax/white-label/integration-pattern.md §2 and compute white_label_active + client_name per that section. Apply §5 (Branding Header) and §7 (Provenance) when composing the Output Format.

Branding Header (only if white_label_active AND client_name != "") — single line at the very top of the rendered output: **<client_name>** Soros-style regime view. Logo handling per integration-pattern.md §5.
Provenance (always present): one line stating branding state per integration-pattern.md §7. If a logo was skipped, append Logo on file: <basename> as a second Provenance line.

AI-interaction disclosure (required regardless of view state): Render parallax-conventions.md §9.2 immediately above the disclaimer below. The persona-specific disclaimer in the output example characterizes the source of the framing; the §9.2 banner characterizes the LLM-generated synthesis itself.

Render the standard disclaimer verbatim from parallax-conventions.md §9.1.

Graceful fallback

If list_macro_countries fails, derive covered markets from RIC suffix defaults per parallax-conventions.md §6 and note degradation.

If macro_analyst fails for a subset of markets, proceed with remaining markets as long as ≥2 succeeded.

If get_telemetry fails (e.g., "Admin org not configured"), single-ticker mode cannot perform Channel B. Fall back to industry-exposure-only evaluation on Channel A. In this fallback state, maximum verdict is partial_match — a single-channel FLAGGED result can never be match because match requires BOTH channels. Note "Channel B: UNAVAILABLE (telemetry env)" in the output.

If build_stock_universe returns no names or times out for any theme: for basket mode, report themes without rankings. For single-ticker mode, Channel A sub-path A1 (universe membership) is NOT_FLAGGED, but sub-path A2 (sector/industry classification match) must STILL be evaluated independently from get_company_info data. Channel A overall is NOT_FLAGGED only if A2 also fails to match. This is the critical non-regression: a build_stock_universe timeout does NOT automatically collapse Channel A to NOT_FLAGGED.