earnings-analysis

earnings-analysis

Purpose

Single-responsibility skill: turn a quarterly earnings event into a structured signal. The output feeds news-intelligence, bull-bear-thesis, and report-composer.

Inputs

• ticker (required).
• Optional: quarter (e.g. Q1-FY26). Defaults to the latest reported.

Methodology

1. Fetch artefacts.
   • 8-K filing containing the earnings press release (US issuers) — SEC EDGAR.
   • The earnings press release PDF / HTML from the IR site.
   • The conference-call transcript — first via financial-datasets MCP if available; otherwise via WebFetch of the IR site or a major transcript provider.
   • Consensus estimates — via financial-datasets MCP (analyst estimates endpoint).
2. Compute beats/misses for: revenue, gross margin, operating margin, EPS, guidance vs prior guidance. Express each as actual - consensus and as a percentage.
3. Extract management guidance verbatim. Map each guidance line to the line item on the income statement it constrains.
4. Score tone. From the prepared-remarks transcript, classify each forward statement on a confident / measured / cautious / defensive scale. Cite the exact quote.
5. Q&A signal. Identify the three questions analysts pressed on hardest (multiple analysts asked, or management deflected). These are usually where the next disappointment lives.

Dependencies

• Consumes: company-profile (segment list to map guidance correctly).
• MCP: financial-datasets (analyst estimates, transcripts where available).
• Web: WebFetch for the 8-K and IR-site artefacts as fallback.

Output schema

JSON at ${CLAUDE_PLUGIN_DATA}/companies/<TICKER>/earnings/<QUARTER>.json:

{
  "quarter": "Q1-FY26",
  "reported_on": "2026-04-30",
  "beats_misses": {"revenue": {"actual_usd_m": ..., "consensus_usd_m": ..., "delta_pct": ..., "source": "..."}},
  "guidance": [{"line": "revenue", "value": "...", "vs_prior": "raised|maintained|lowered", "quote": "...", "source": "..."}],
  "tone": [{"topic": "...", "label": "confident|measured|cautious|defensive", "quote": "...", "source": "..."}],
  "qa_pressure_points": [{"topic": "...", "analysts": ["..."], "deflection": true, "source": "..."}]
}

Source citation policy

Every quote carries a source: pointing to the transcript URL + speaker + approximate timestamp or paragraph. Every consensus figure cites the MCP endpoint and as-of date.