← Back to list
alphaear-predictor Checked
Market prediction skill using Kronos. Use when user needs finance market time-series forecasting …
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-predictor" skill by following the instructions at https://skill123.me/install/alphaear-predictor.
⌨️ Command line install
Run in your terminal — downloads and installs to ~/.claude/skills/.
curl -fsSL https://skill123.me/install/alphaear-predictor.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 · 1.2 MBAbout this skill
Overview
Market prediction skill using Kronos. Use when user needs finance market time-series forecasting or news-aware finance market adjustments.
Source
- Repo: https://github.com/RKiding/Awesome-finance-skills
- Path: skills/alphaear-predictor
Score breakdown
Trigger
5.0
What plus When clause (time-series forecasting / news-aware adjustments); no When-Not; no trigger variants; English-only.
Structure
6.7
Monolithic application bundle (11,236 lines) incl. a binary .pt checkpoint; one useful reference (PROMPTS.md); duplicated '### 1. Forecast Market Trends' heading; minimal disclosure of the code mass.
Workflow
6.4
Two-stage design (deterministic Kronos base forecast -> agentic adjustment with strict-JSON output contract) is sound; guarded model loading with local-cache-first and download fallback; undermined by broken example import path and API mismatch between SKILL.md and code; test tolerates any init failure (weak verification).
Content
7.0
Config section is concrete (env vars, model paths, CAUTION note); example code is broken (wrong module path, non-existent predict() signature); mixed Chinese/English.
Engineering
8.0
Valid frontmatter; ML dependencies and environment variables well declared; referenced paths exist but the documented import path is wrong (scripts.utils.kronos_predictor vs scripts/kronos_predictor).
Security
8.5
Manifest: HuggingFace downloads (NeoQuasar/Kronos-Tokenizer-base, Kronos-base, sentence-transformers/all-MiniLM-L6-v2) — base model download declared in SKILL.md; bundled 1.28MB .pt checkpoint loaded with weights_only=True (good, with explicit security CAUTION); evaluation.py torch.load lacks weights_only (inconsistent); llm/search/news code with provider env keys again bundled beyond declared scope; writes model exports dir. No injection, no hardcoded secrets, no persistence. No investment-advice disclaimer for a market-prediction skill (minor compliance deduction).
Kronos-based K-line forecasting with an agentic news-adjustment prompt. Notable security awareness: weights_only=True on torch.load with an explicit CAUTION note about untrusted checkpoints, env-var config, well-declared ML deps. But: duplicated heading, example imports scripts.utils.kronos_predictor while the file lives at scripts/kronos_predictor.py, API mismatch (predict() vs get_base_forecast()), utils/predictor/evaluation.py loads checkpoints WITHOUT weights_only, and the whole 11k-line FinAgent bundle (news/search/LLM routing) ships again behind a prediction-only description.