← Back to list
alphaear-sentiment Checked
Analyze finance text sentiment using FinBERT or LLM. Use when the user needs to determine the sen…
6.8Overall rating
Installation
🤖 Install via AI
Copy the prompt below and send it to your AI assistant (e.g. Claude Code) — it will follow the instructions and install automatically.
Install the "alphaear-sentiment" skill by following the instructions at https://skill123.me/install/alphaear-sentiment.
⌨️ Command line install
Run in your terminal — downloads and installs to ~/.claude/skills/.
curl -fsSL https://skill123.me/install/alphaear-sentiment.sh | bash
📦 Download ZIP
Download the zip and extract it into your skills directory (e.g. ~/.claude/skills/), then restart your session.
⬇ Download v1.0.0 · 15 KBAbout this skill
Overview
Analyze finance text sentiment using FinBERT or LLM. Use when the user needs to determine the sentiment (positive/negative/neutral) and score of financial text markets.
Source
- Repo: https://github.com/RKiding/Awesome-finance-skills
- Path: skills/alphaear-sentiment
Score breakdown
Trigger
5.0
What plus explicit 'Use when the user needs to determine sentiment...' clause; no When-Not; no trigger variants; description in English only though the analysis prompt is Chinese.
Structure
7.3
57-line SKILL.md; scripts/ bundles a 581-line shared database_manager and an unused llm/ factory tree; no references/; progressive disclosure minimal.
Workflow
6.4
Dual-path design is sound: deterministic BERT pipeline for speed, agent-executed LLM prompt with a concrete scoring guide (-1..1 bands) and DB write-back helper; local-cache-first model loading with download fallback; batch mode for unanalyzed news rows.
Content
7.0
Mostly clear; duplicated '## Capabilities' heading; no runnable CLI example (API-level usage only); mixed Chinese/English docs.
Engineering
8.0
Valid frontmatter; deps list torch/transformers/sqlite3 but loguru undeclared; modules lack shebangs (importable); referenced paths exist; init test avoids model download via mode='llm' and in-memory DB.
Security
8.0
Manifest: env reads SENTIMENT_MODE/BERT_SENTIMENT_MODEL; network: HuggingFace model download (uer/roberta-base-finetuned-chinanews-chinese) on first use — undeclared in SKILL.md; llm/factory.py carries env-key LLM provider configs (DashScope/z.ai/OpenRouter/DeepSeek) but is not invoked by this skill's tools (dead code); writes only local sqlite db. No hardcoded secrets, no injection, no persistence. Deductions: undeclared HF egress, dead egress-capable code, and FinBERT-vs-actual-model naming mismatch.
Two-mode sentiment skill (local HF pipeline + agentic LLM prompt with scoring guide) with DB persistence and a hermetic init test. Issues: duplicated '## Capabilities' heading; description says 'FinBERT' but the default model is uer/roberta-base-finetuned-chinanews-chinese (a news classifier, not FinBERT); undeclared HuggingFace model download; bundled llm/ factory (DashScope/z.ai/OpenRouter/DeepSeek) is dead code for this skill's entry point.