Skillv3.0.0

ClawScan security

天气套利助手 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousMar 5, 2026, 8:30 AM

Verdict: suspicious
Confidence: medium
Model: gpt-5-mini
Summary: Code and instructions mostly match a weather-arbitrage scanner, but there are notable mismatches and unknowns (payment/skillpay module, omitted files, and the skill claims live profitability while code appears to only generate signals), so review before running or entrusting funds.
Guidance: What to check before installing or running: 1) Inspect scripts/skillpay.js and any omitted files for hardcoded endpoints, API keys, or code that sends data to third-party servers — the provided package uses chargeUser() and may redirect to payment URLs. 2) Confirm there is no code that will place real trades or require/exfiltrate exchange credentials — the visible code only generates signals and simulations, so treat claims of live trading performance as unverified marketing. 3) Run the skill in a sandboxed environment and monitor outbound network requests (especially to unknown domains) before providing real funds. 4) If you plan to use any paid features, verify the payment flow and merchant identity; avoid entering sensitive credentials until reviewed. 5) Ask the author for source provenance, a homepage, and an explanation of where live trade execution occurs (if any). If you want a higher-confidence security assessment, provide the full contents of the omitted files (including scripts/skillpay.js) so they can be reviewed for hidden endpoints or credential handling.

Review Dimensions

Purpose & Capability: concernThe skill claims automated, high-return live trading (91% win rate, large monthly profit) but the visible code generates signals and simulations and does not show any market-order placement or account credential handling for executing trades. Claiming live trading performance while shipping only scanners/simulators is an incoherence that could be misleading to non-technical users.
Instruction Scope: noteSKILL.md instructs running node scripts which the code implements. The runtime code performs network calls to public APIs (NOAA, Polymarket, Kalshi) and uses setInterval for continuous scanning. It does not prompt to read unrelated host files or environment variables, but it calls chargeUser() (skillpay) which may contact external payment endpoints — that module's behavior was not provided in full in the supplied text and could change risk.
Install Mechanism: okNo install spec; package.json lists only a runtime dependency (node-fetch). There are no downloads from unknown hosts or archive extraction steps. Running requires Node and network access.
Credentials: noteDeclared requirements list no environment variables or credentials, which fits the code that only reads SKILLPAY_DEV. However the presence of a 'skillpay' payment/charging module suggests external payment flows; lack of declared credentials for payment processors or exchange APIs is notable and should be inspected (payment may be handled via returned URLs or third-party pages).
Persistence & Privilege: okSkill does not request always:true and contains no code that modifies other skills or global agent configuration. It sets local intervals for continuous scanning when run, which is expected for a monitoring tool.