earnings-call-processor

by terrycarter1985

Install this Skill

curl -fsSL -o earnings-call-processor.zip https://skill.seembox.com/api/v1/skills/earnings-call-processor/download && unzip earnings-call-processor.zip

Download earnings-call-processor.zip

SKILL.md About

---
name: earnings-call-processor
description: Automated earnings call transcript processing and analysis. Use when the user provides an earnings call audio file or transcript and wants structured analysis including key metrics, management sentiment, forward guidance, and stock data correlation. Triggers on phrases like "process earnings call", "analyze earnings call", "财报电话会", "财报分析", "处理财报", or when an audio/transcript file related to quarterly earnings is provided. Outputs structured analysis as a Feishu draft document. Fully automated pipeline: audio → transcript → stock data → structured analysis → Feishu draft.
version: 1.1.0
---

# Earnings Call Processor v1.1.0

Fully automated pipeline: audio → transcript → stock data → structured analysis → Feishu draft.

## Quick Start

Provide an audio file or transcript with ticker and quarter:

```
User: "处理这份AAPL Q1 2025财报电话会 audio.mp3"
User: "分析这份MSFT Q4 2024财报" (with transcript file)
```

Agent executes the full pipeline automatically—no manual step-by-step.

## Automated Workflow

### 1. Run Pipeline Script

```bash
python3 scripts/process_call.py --ticker {TICKER} --quarter "{QUARTER}" --audio {path}
# or
python3 scripts/process_call.py --ticker {TICKER} --quarter "{QUARTER}" --transcript {path}
```

Script outputs JSON with transcript, stock data, and next-step instructions.

If `--audio` is given, `scripts/transcribe.py` runs automatically (requires faster-whisper).
Stock data is fetched via yfinance (price, 5d history, company info).

### 2. Apply Analysis Framework

Read `references/analysis_framework.md` and apply it to the transcript + stock data from step 1.

Generate structured markdown with all 8 required sections:
1. 执行摘要 (beat/miss/meet, headline numbers, market reaction)
2. 核心财务指标 (revenue, EPS, margins with YoY and vs-consensus)
3. 管理层基调 (sentiment: 🟢/🟡/🔴 with quoted evidence)
4. 战略亮点 (M&A, products, expansion, capital allocation)
5. Q&A 重点 (top 3-5 analyst questions + responses)
6. 前瞻指引 (raised/lowered/maintained guidance)
7. 股价关联 (after-hours reaction, volume, sector comparison)
8. 投资要点 (bull/bear/risk/watch)

### 3. Create Feishu Draft

```bash
python3 .claude/skills/feishu-doc/scripts/doc_ctl.py create "财报分析: {COMPANY} {QUARTER}" --content "<markdown>"
```

Title format: `财报分析: {公司名} {YEAR} Q{N}`

### 4. Notify User

After draft creation, send the user a brief summary with the Feishu doc link:
- Company + quarter
- Beat/miss verdict
- Key metric highlights
- Sentiment classification
- Doc link

## Transcription (Audio Input)

```bash
python3 scripts/transcribe.py <audio_file> [--model base] [--language en] [--output out.txt]
```

Uses faster-whisper (CPU, int8). Models: tiny/base/small/medium/large.

## Stock Data Enrichment

The pipeline script auto-fetches via yfinance:
- Current price, previous close, market cap, PE ratios, 52-week range
- 5-day price history (earnings window)
- Company name, sector, industry

Fallback: use `finance` skill tools (`yfinance_get_stock_price`, `yfinance_get_historical_data`, `yfinance_get_stock_info`) if inline yfinance fails.

## Configuration

| Env Var | Default | Description |
|---------|---------|-------------|
| `WHISPER_MODEL` | `base` | faster-whisper model size |
| `WHISPER_DEVICE` | `cpu` | cpu or cuda |

## v1.1.0 Changelog

- End-to-end pipeline: `process_call.py` now fetches stock data automatically
- Auto-notify user after Feishu draft creation
- Structured JSON output with analysis instructions
- Graceful fallback when yfinance unavailable
- Version field in frontmatter
- Improved analysis framework with investment takeaways

_meta.json Metadata

{
  "ownerId": "kn72jmd20ws94jr0p8b3p24wyn82abgj",
  "publishedAt": 1778146435915,
  "slug": "earnings-call-processor",
  "version": "2.1.0"
}