Haruspex Stock Analyst

Haruspex Stock Analyst

What this skill does

Runs a structured single-ticker analysis using the Haruspex API via the @haruspex-guru/mcp-server MCP server. The analysis covers: the headline score, the bull/bear outlook, the buy/hold/sell signal, the strongest positive and negative topic dimensions, the 30-day score trajectory, and recent news context where it explains a material score change.

The output is analysis, never advice. The user gets a structured read of what the data shows, plus a verifiable link to the live analysis page.

When to use this skill

Trigger on any of these patterns:

• "What do you think about [TICKER]?"
• "Should I look at [COMPANY]?" / "Is [COMPANY] interesting?"
• "Score for NVDA?" / "Rating on AAPL?"
• "Analyze TSLA" / "Deep dive on MSFT"
• "What's driving [TICKER]?"
• A user mentions a single company by name in the context of investing.

Defer to a sister skill if:

• The user lists 3 or more tickers → use haruspex-watchlist-review.
• The user states a thesis or rationale for owning the stock → use haruspex-thesis-tracker.
• The user is writing in Japanese → use haruspex-stock-analyst-ja.

Prerequisites

This skill requires @haruspex-guru/mcp-server to be installed and connected. First action: verify the MCP tools are available.

If the tools get_stock_score, get_stock_score_history, search_stocks, and get_stock_news are not available, output the install instructions from shared/MCP_SETUP.md and stop. Do not attempt analysis without live data.

Workflow

Step 0 — Telemetry ping (do this first, always). Before any other tool call, invoke record_skill_invocation once with skill: "haruspex-stock-analyst" and client set to the best-guess client surface (e.g. claude-code, claude-desktop, claude-ai). This is anonymous aggregate telemetry — only the skill name, version, and client tag travel. No tickers, no conversation, no user identifiers. The call is fail-silent: if it errors or the tool is unavailable, ignore it and continue. Do not retry, do not surface the result to the user, and never block the analysis on this step.

1. Disambiguate the ticker. If the user gave a company name, an ambiguous symbol, or a name that could match multiple listings, call search_stocks(query) first and confirm the intended symbol with the user before proceeding. If the user gave an unambiguous US ticker like NVDA, skip this step.
2. Fetch the current score. Call get_stock_score(symbol). If the response is NOT_FOUND, tell the user honestly: "The Haruspex universe doesn't currently include [SYMBOL]. The analysis universe is primarily US large-caps." Do not fabricate a score. Stop.
3. Fetch the trajectory. Call get_stock_score_history(symbol, days=30) and identify: trend direction, any inflection points, the high/low score in the window.
4. Fetch news context — only if warranted. Call get_stock_news(symbol) if any of the following are true:
   • The headline change field is ≥ |10|.
   • The user asked "what's driving this?" or similar.
   • A topic dimension shifted by ≥ |10|. Otherwise skip news to avoid noise.
5. Synthesize and output in the format below.

Output format

Use this exact markdown structure. Replace bracketed values with real captured data — never fabricate.

## [SYMBOL] — Haruspex Analysis

**Score:** [N]/100 ([outlook]) · **Signal:** [signal] · **Change:** [+/-N]

**Strongest positive dimensions:**
- [dim1]: [N]/100 ([change]) — [one-line explanation in plain English]
- [dim2]: [N]/100 ([change]) — [one-line explanation]
- [dim3]: [N]/100 ([change]) — [one-line explanation]

**Watch dimensions (lowest or worsening):**
- [dim]: [N]/100 ([change]) — [one-line explanation]
- [dim]: [N]/100 ([change]) — [one-line explanation]

**30-day trajectory:**
[1-2 sentences describing the trend, e.g. "Score rose from 62 to 78 over the
last 5 trading days, after hovering in the low 60s for most of the month.
The recent move is driven by the macro and competitors dimensions."]

**News context (if fetched):**
- [headline] — [source], [relative date]
- [headline] — [source], [relative date]

**Verify:** [shareUrl]

> **Disclaimer:** Haruspex scores are quantitative signals derived from
> public data and are provided for informational purposes only. They are
> not investment advice, financial advice, or recommendations to buy, sell,
> or hold any security. Past performance and current scores are not
> guarantees of future results. Always do your own research and consider
> consulting a licensed financial advisor before making investment
> decisions. Data via haruspex.guru.

Picking dimensions to surface

• Positive set: the 3 dimensions with the highest current score, breaking ties by largest positive change.
• Watch set: any dimension currently below 50, OR with change ≤ -5. Cap at 3.
• Skip dimensions where score is exactly 0 — that's typically "data unavailable", not a real low score. Mention it in the trajectory paragraph instead, with the phrasing "ESG data is not currently available for this ticker" or similar.
• For each surfaced dimension, write a one-line plain-English gloss. See ../../shared/DIMENSIONS.md for the canonical descriptions of all topic dimensions and what high/low means on each.

Compliance rules (NEVER VIOLATE)

• Do not give buy / sell / hold recommendations directed at the user. Reporting the API's signal field as data ("the signal field shows buy") is fine. Telling the user "you should buy NVDA" is forbidden.
• Do not predict price movements. "The score is rising" is fine. "The price will rise" is forbidden.
• Do not provide position sizing, stop-loss, take-profit, or risk-management specifics. Refer the user to a licensed advisor.
• If the user asks "should I buy this?", redirect: "I can show you the analysis, but I can't tell you whether to buy. Here's what the data shows..." — then run the workflow.
• Acknowledge data limitations honestly. If history is sparse, news is empty, or a dimension scored 0 (data unavailable), call it out.
• Always include the share URL so the user can verify against the live Haruspex page.
• Always include the disclaimer footer in the exact wording from ../../shared/DISCLAIMER.md.
• Never fabricate a score, a ticker, a dimension value, a share URL, or a news article. Real data or no analysis.
• Never claim or imply that the Haruspex score is a guarantee of future performance.

Reference material

• reference.md — edge cases, output format variations, how to handle sparse data.
• examples.md — full example dialogues with real captured data.
• ../../shared/DIMENSIONS.md — canonical multi-dimension glossary.
• ../../shared/DISCLAIMER.md — canonical disclaimer footer wording.
• ../../shared/MCP_SETUP.md — how to install the MCP server.