Description

训练动态知识图谱嵌入模型，学习时序实体关系表示，支持链接预测和时间预测任务。

README (SKILL.md)

金融知识图谱嵌入 (finance-kg-embedding)

Name: Finance Kg Embedding
Author: tangweigang-jpg

训练动态知识图谱嵌入模型，学习时序实体关系表示，支持链接预测和时间预测任务。

Pipeline

data_collection -> data_storage -> factor_computation -> target_selection -> trading_execution -> visualization

Top Use Cases (5 total)

KGTransformer Model Training Pipeline (`UC-101`)

Training a knowledge graph-based transformer model for temporal/dynamic knowledge graph embedding tasks to learn entity and relation representations o Triggers: training, knowledge graph, KGTransformer

Dynamic Knowledge Graph Model Training (`UC-102`)

Training dynamic knowledge graph models to learn temporal entity and relation embeddings for link prediction and event time prediction tasks Triggers: knowledge graph, dynamic graph, temporal modeling

Early Stopping Training Utility (`UC-103`)

Preventing overfitting during model training by automatically stopping training when validation performance stops improving, with checkpoint managemen Triggers: early stopping, overfitting prevention, model training

For all 5 use cases, see references/USE_CASES.md.

Execute trigger: When user intent matches intent_router.uc_entries[].positive_terms AND user uses action verb (run/execute/跑/执行/backtest/fetch/collect)

What I'll Ask You

Target market: A-share (default), HK, or crypto? (US stocks in ZVT are half-baked — stockus_nasdaq_AAPL exists but coverage is thin)
Data source / provider: eastmoney (free, no account), joinquant (account+paid), baostock (free, good history), akshare, or qmt (broker)?
Strategy type: MACD golden-cross, MA crossover, volume breakout, fundamental screen, or custom factor?
Time range: start_timestamp and end_timestamp for backtest period
Target entity IDs: specific stocks (stock_sh_600000) or index components (SZ1000)?

Semantic Locks (Fatal)

ID	Rule	On Violation
`SL-01`	Execute sell orders before buy orders in every trading cycle	halt
`SL-02`	Trading signals MUST use next-bar execution (no look-ahead)	halt
`SL-03`	Entity IDs MUST follow format entity_type_exchange_code	halt
`SL-04`	DataFrame index MUST be MultiIndex (entity_id, timestamp)	halt
`SL-05`	TradingSignal MUST have EXACTLY ONE of: position_pct, order_money, order_amount	halt
`SL-06`	filter_result column semantics: True=BUY, False=SELL, None/NaN=NO ACTION	halt
`SL-07`	Transformer MUST run BEFORE Accumulator in factor pipeline	halt
`SL-08`	MACD parameters locked: fast=12, slow=26, signal=9	halt

Full lock definitions: references/LOCKS.md

Top Anti-Patterns (14 total)

AP-MACRO-DATA-001: SEC EDGAR Rate Limit Violation
AP-MACRO-DATA-002: Temporal Knowledge Graph Look-Ahead Bias
AP-MACRO-DATA-003: Technical Indicator Look-Ahead Bias via Missing Shift

All 14 anti-patterns: references/ANTI_PATTERNS.md

Evidence Quality Notice

[QUALITY NOTICE] This crystal was compiled from blueprint finance-bp-080. Evidence verify ratio = 19.0% and audit fail total = 15. Generated results may have uncaptured requirement gaps. Verify critical decisions against source files (LATEST.yaml / LATEST.jsonl).

Reference Files

File	Contents	When to Load
references/seed.yaml	V6+ 全量权威 (source-of-truth)	有行为/决策争议时必读
references/ANTI_PATTERNS.md	14 条跨项目反模式	开始实现前
references/WISDOM.md	跨项目精华借鉴	架构决策时
references/CONSTRAINTS.md	domain + fatal 约束	规则冲突时
references/USE_CASES.md	全量 KUC-* 业务场景	需要完整示例时
references/LOCKS.md	SL-* + preconditions + hints	生成回测/交易代码前
references/COMPONENTS.md	AST 组件地图（按 module 拆分）	查 API 时

Compiled by Doramagic crystal-compilation-v6.1 from finance-bp-080 blueprint at 2026-04-22T13:00:31.071227+00:00. See human_summary.md for non-technical overview.

Usage Guidance

This skill looks like a legitimate finance ML blueprint, but it contains runtime instructions that will run host-level Python checks, install or require packages (zvt, Python 3.12+, 'uv' package manager), check and write to ZVT_HOME (~/.zvt), and includes trading execution steps — none of which are declared in the registry metadata. Before installing or invoking it: 1) Ask the author to provide a clear install spec and a list of required binaries/env vars (e.g., Python version, ZVT_HOME, broker/API keys) and to confirm whether the skill will execute trades autonomously. 2) Do not provide broker or provider credentials until you verify where/how they are used; prefer ephemeral/test credentials. 3) Run the skill in an isolated environment (sandbox or VM) first, and inspect any pip installs (verify package sources). 4) If you plan to allow it to execute commands on your host, review the seed.yaml and SKILL.md preconditions and confirm you accept the described behavior. 5) If you need lower risk, request a version that is purely read-only (no precondition shell/python commands and no trading_execution) or with explicit explicit declarations of all required env vars and install steps.

Capability Analysis

Type: OpenClaw Skill Name: finance-kg-embedding Version: 0.3.3 The finance-kg-embedding skill bundle is a comprehensive framework for training and evaluating dynamic knowledge graph models for financial applications using the zvt library. The bundle contains extensive documentation, including 'Semantic Locks' (e.g., SL-01 to SL-12) and 'Fatal Constraints' (e.g., finance-C-001) specifically designed to prevent logical errors and financial modeling pitfalls like look-ahead bias and improper data splitting. No evidence of malicious intent, data exfiltration, or harmful prompt injection was found; the instructions and execution protocols (seed.yaml) are strictly aligned with the stated purpose of quantitative financial analysis and model training.

Capability Tags

crypto

Capability Assessment

ℹ Purpose & Capability

The name/description (dynamic finance KG embedding, link/time prediction) match the SKILL.md content and reference components. However SKILL.md also describes a broader end-to-end pipeline (including data collection, recorder usage and 'trading_execution') and requires Python 3.12+ with 'uv' package manager in the compatibility block — none of these runtime requirements are reflected in the registry metadata (which lists no required binaries/env/config). That discrepancy is unexpected and should be clarified.

⚠ Instruction Scope

SKILL.md and seed.yaml include explicit preconditions and execution steps that run host-level python commands and check/modify host state: e.g., PC-01..PC-04 run python -c checks for zvt, instruct pip install zvt on failure, check/write permissions for ZVT_HOME (~/.zvt), and recommend running recorders. seed.yaml's execution_protocol instructs agents to re-load seed.yaml before behavioral decisions and to run install_recipes[] on the host. Those are host-facing operations that go beyond simply answering ML questions and could cause the agent to execute arbitrary shell/python commands and touch filesystem paths not declared in the skill metadata.

ℹ Install Mechanism

There is no declared install spec (instruction-only), which is lower risk. But seed.yaml/execution_protocol and SKILL.md reference installing packages (pip install zvt, and host_adapter.install_recipes[]) and require Python 3.12+ with an 'uv' package manager. Because the registry shows no install steps, there's a mismatch: the skill may expect to trigger installs at runtime even though none are declared in metadata. This should be clarified before running.

⚠ Credentials

Registry metadata lists no required env vars, yet SKILL.md/seed.yaml reference ZVT_HOME, running recorders that contact data providers (eastmoney/joinquant/akshare/qmt), and trading execution semantics (semantic locks). The skill may therefore attempt to read/write ~/.zvt, access provider APIs, or invoke trading flows that require broker credentials — none of which are declared or scoped. The absence of declared credentials or config paths while instructions expect host-side state is disproportionate.

ℹ Persistence & Privilege

The skill does not request always:true and is user-invocable (normal). However seed.yaml's execution_protocol directs agents to re-read seed.yaml for any behavioral decision and to run preconditions/install triggers; that practice broadens the operational surface by repeatedly invoking host checks and possible installs. It's not an outright privilege escalation but is noteworthy and should be understood by the user.

Version History

v0.3.3

v0.3.3: bilingual metadata injected. H1 shows 金融知识图谱嵌入; tagline replaced with skill-specific Chinese hook; tags upgraded to Level 1-4.

v0.3.1

Remove install.sh — knowledge-only bundle. Host AI consumes directly from URL; no user-side installation needed. Fixes ClawHub suspicious flag.

v0.3.0

Doramagic crystal portfolio v0.3.0. Full 5-layer bp-009 standard. github.com/tangweigang-jpg/doramagic-skills

Metadata

Slug finance-kg-embedding

Version 0.3.3

License MIT-0

All-time Installs 0

Active Installs 0

Total Versions 3

Frequently Asked Questions

What is Finance Kg Embedding?

训练动态知识图谱嵌入模型，学习时序实体关系表示，支持链接预测和时间预测任务。 It is an AI Agent Skill for Claude Code / OpenClaw, with 109 downloads so far.

How do I install Finance Kg Embedding?

Run "/install finance-kg-embedding" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Finance Kg Embedding free?

Yes, Finance Kg Embedding is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Finance Kg Embedding support?

Finance Kg Embedding is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Finance Kg Embedding?

It is built and maintained by Tang Weigang (@tangweigang-jpg); the current version is v0.3.3.

More Skills

Finance Kg Embedding