Skillv1.0.3

ClawScan security

moss-trade-bot-factory-en · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 19, 2026, 4:14 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's code, instructions, and requirements are internally consistent with a local crypto bot/backtest factory; network access and credential use are limited to explicit platform operations and the bundled scripts.
Guidance: This package appears to be what it says: a local trading-bot/backtest tool with optional platform upload/trading. Before installing or running it: 1) Run it in an isolated environment (VM/container) or a throwaway account; setup_env.py creates a .venv in the skill folder—avoid installing globally. 2) Do not place real exchange API keys or funds-level credentials at the default creds path unless you intend to enable live/upload features; use a test/simulated key. 3) Verify the platform URL before uploading (default: https://ai.moss.site) and only provide credentials when you explicitly ask the agent to bind/upload/trade. 4) Expect the skill to contact Binance (via the fetcher) if you allow data downloads. 5) If you want stricter interaction, instruct the agent to pause for confirmations at additional points (SKILL.md currently allows many automated steps except specified checkpoints). If you need more confidence, request a security review of the core.fetcher and any network-facing code used to download market data.

Review Dimensions

Purpose & Capability: okName/description match the implementation: Python scripts create params, run backtests, evolve strategies, and optionally upload/create live bots against a platform URL. Required binary is only python3 and bundled dependencies (pandas, numpy, ccxt, scipy) are appropriate for the stated functionality.
Instruction Scope: noteSKILL.md directs the agent to run the included scripts, build a local venv, use bundled CSV or fetch Binance UM data, and only ask specific confirmations (install venv, first backtest, evolution enable, first live switch, each manual order). It references a local creds file (~/.moss-trade-bot/agent_creds.json) and /tmp intermediates; these are used only for explicit platform/upload/live flows. The instruction to 'follow steps strictly' and to auto-fill many settings (ask only one question) grants the agent operational autonomy within the local workflow—this is coherent but worth noting to users who want more interactive confirmations.
Install Mechanism: okThere is no external download/install spec in the registry; the skill includes a setup_env.py that creates a local virtualenv and installs only the bundled requirements.txt. No remote or obfuscated installers are used in the provided files.
Credentials: noteThe skill declares no required environment variables and does not attempt to read hidden env vars. Platform credentials are read from a local JSON creds file (path configurable, default ~/, e.g. ~/.moss-trade-bot/agent_creds.json) only when the user asks to upload/bind/trade live. This is proportional, but users should be aware that providing that local creds file (or leaving sensitive keys at the default path) enables networked trading and verification flows.
Persistence & Privilege: okSkill does not request always:true and does not modify other skills. It runs locally and only stores/reads its own cred file when platform features are explicitly used.