Credit Scorecard

基于监督学习、决策树或聚类等多种算法，自动为评分卡变量生成最优分箱边界，同时支持单调性约束和缺失值处理。

Security Scan

The skill's description (credit scorecard / bucketing) is plausible, but the runtime instructions mix in trading/backtest workflows, require reading host workspace files and running Python/pip precondition commands (including recommending pip install), which is disproportionate and potentially risky; proceed with caution and inspect before use.

Purpose & Capability

Name/description focus on credit-scorecard bucketing and WoE pipelines, but SKILL.md and human_summary heavily reference trading/backtest (ZVT, MACD, trading execution, backtests) and semantic locks for trading. The skill appears to conflate credit-risk bucketing and a trading/backtest toolkit (ZVT). This mismatch between declared purpose (scorecard) and the broader finance/backtesting capabilities is unexpected and worth verifying with the author.

Instruction Scope

SKILL.md contains explicit runtime protocol: agents MUST re-read seed.yaml, run precondition checks that execute python -c commands, and follow an execute_protocol that may cause the agent to run installation or remediation commands (e.g., pip install zvt, zvt.init_dirs). The skill also references host workspace paths (placeholders like {host_workspace}/skills/), and enforces re-reading local seed.yaml before behavioral decisions. Those instructions allow the agent to read and interact with local files and to attempt package installation—scope creep beyond a simple 'instruction-only' helper.

Install Mechanism

There is no formal install spec (instruction-only, no code files), which is low disk persistence risk. However SKILL.md and seed.yaml explicitly expect Python 3.12+ and an 'uv' package manager and include preconditions that remediate by running pip install zvt. Although not an embedded installer, the instructions encourage installing third-party packages at runtime—which can pull arbitrary code from PyPI and should be treated as an install action.

Credentials

The skill declares no required env vars or credentials, which is good. But SKILL.md references ZVT_HOME and includes precondition scripts that create/write to ~/.zvt (touch/unlink), and workspace_resolution points to host workspace paths. The skill therefore expects file-system access and specific environment layout; it does not request secrets but will read/write config directories and may install packages. This level of access is broader than implied by a simple 'bucketing' helper.

Persistence & Privilege

always:false and no install spec mean the skill is not requesting permanent automatic inclusion or explicit system-level persistence. Autonomous invocation (disable-model-invocation:false) is the platform default; by itself it is not flagged. Note: combined with instruction_scope concerns (ability to read workspace and run installs), autonomous invocation would increase potential impact—consider restricting automatic runs if you don't fully trust the skill.

What to consider before installing

This is an instruction-only skill (no code files), but its instructions go beyond simple bucketing: it intermixes credit-scorecard logic with trading/backtest workflows (ZVT, MACD), asks agents to re-read local seed.yaml, run python checks and even suggests pip installing zvt and initializing ~/.zvt. Before installing or enabling this skill: 1) Inspect seed.yaml and the referenced files yourself to confirm they don't reference sensitive host paths or run unexpected commands. 2) Do not grant it autonomous execution in a production agent until you've sandbox-tested it (use an isolated VM/container). 3) If you need zvt, prefer installing packages manually from a vetted source and verify package integrity (pinned versions, known PyPI authors). 4) Be cautious about allowing the skill to read your host workspace or write to ~/.zvt; if possible, run it with a restricted working directory. 5) If the skill is intended only for credit scoring, ask the author to remove trading/backtest-specific preconditions and workspace-resolution rules or to provide a dedicated, minimal variant. If you lack the ability to audit the files, treat this skill as untrusted and avoid enabling autonomous invocation.

信用评分卡 (credit-scorecard)

基于监督学习、决策树或聚类等多种算法，自动为评分卡变量生成最优分箱边界，同时支持单调性约束和缺失值处理。

Pipeline

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

Top Use Cases (43 total)

Optimal Supervised Bucketing (UC-1)

Automatically find optimal bucket boundaries that maximize predictive power while respecting monotonicity constraints Triggers: optimal, supervised, monotonic

Decision Tree Supervised Bucketing (UC-2)

Use supervised learning to find bucket boundaries based on target variable correlation Triggers: decision tree, supervised, pre-bin

Equal Width Unsupervised Bucketing (UC-3)

Divide numerical features into N equally spaced intervals regardless of data distribution Triggers: equal width, unsupervised, histogram

For all 43 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-CREDIT-RISK-001: Empty DataFrame passed to bucketing pipeline
• AP-CREDIT-RISK-002: Multi-dimensional target array causing WoE shape mismatch
• AP-CREDIT-RISK-003: OptimalBucketer receiving high-cardinality numerical features

All 14 anti-patterns: references/ANTI_PATTERNS.md

Evidence Quality Notice

[QUALITY NOTICE] This crystal was compiled from blueprint finance-bp-050. Evidence verify ratio = 78.6% and audit fail total = 24. 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-050 blueprint at 2026-04-22T13:00:17.518473+00:00. See human_summary.md for non-technical overview.