Ledger Plaintext Accounting
WarnAudited by ClawScan on May 10, 2026.
Overview
The skill is labeled as plaintext accounting, but its instructions shift into stock/crypto quant trading, broker use, and order execution without clear live-trading safeguards.
Review this skill carefully before installing. Use it only if you intend quant trading/backtesting, keep broker integrations disabled or read-only by default, run any Python/ZVT setup in an isolated environment, and require explicit confirmation before any live financial action.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A user expecting accounting help could be routed into trading-strategy or order-execution workflows.
The skill is presented as a plaintext ledger/accounting engine, but the main workflow is a quant trading pipeline ending in trading execution. That mismatch can cause users to trust it for accounting while it steers the agent toward a different, higher-impact domain.
description: 通过字节码驱动的复式记账引擎... Pipeline: data_collection -> data_storage -> factor_computation -> target_selection -> trading_execution -> visualization
Rename and describe the skill as quant trading/backtesting if that is the real purpose, or remove trading/broker instructions from an accounting skill.
If connected to a broker or trading environment, the agent could generate or run workflows that affect real positions or money.
The instructions contemplate broker-backed workflows and order sizing, but the artifacts do not clearly bound these actions to simulation/backtesting or require a final user approval before live financial actions.
Data source / provider: eastmoney..., joinquant (account+paid)... or qmt (broker)?; SL-01: Execute sell orders before buy orders in every trading cycle; SL-05: TradingSignal MUST have EXACTLY ONE of: position_pct, order_money, order_amount
Require an explicit backtest-vs-live mode, default to read-only/backtest, and require user confirmation for every live order or broker mutation.
Users may be asked to connect financial accounts or broker tools without a clear permission boundary.
Paid-provider and broker access implies user account authority, but the provided metadata declares no primary credential, required environment variables, or config paths, and the docs do not specify scopes or read-only vs trading permissions.
Data source / provider: eastmoney (free, no account), joinquant (account+paid), baostock..., akshare, or qmt (broker)?
Declare all credential/config requirements and state exactly whether each integration is read-only, paid-data access, paper trading, or live trading.
Installing the skill's recommended dependency can modify the local Python environment and pull code from an external package source.
The skill has no install spec, but its reference instructions tell the user/agent to install an external Python package and initialize it. This is purpose-aligned for ZVT workflows, but it is under-declared and not pinned in the artifacts.
PC-01: python3 -c 'import zvt; print(zvt.__version__)' → on_fail: Run: python3 -m pip install zvt then re-run: python3 -m zvt.init_dirs
Install ZVT only in an isolated environment, pin versions, and require user approval before running package installation commands.
The setup may run local Python commands and create or modify files in the user's ZVT data directory.
The preconditions include running Python modules and writing a test file under ZVT_HOME. That is plausible for market-data setup, but it is local code execution and filesystem mutation that users should approve.
PC-02: Run recorder first: python3 -m zvt.recorders.em.em_stock_kdata_recorder --entity_ids stock_sh_600000; PC-04: test_f.touch(); test_f.unlink()
Run these commands manually or only after confirmation, and set ZVT_HOME to a dedicated directory if using the skill.