Crab Catch

SUSPICIOUS: the skill’s browsing and extraction behavior broadly fits its stated purpose, and the install path appears to be the official project workflow, but it grants an agent powerful interaction over arbitrary web content. The main risk is indirect prompt injection and autonomous website actions, not clear malware or credential theft.

SUSPICIOUS. The overall research purpose broadly matches the data sources, but the footprint is too expansive and trust-heavy: mandatory startup execution, centralized third-party API routing with auth headers, and recursive processing of untrusted web/social content using browser and file-write capabilities. The main concern is data-flow and prompt-injection risk rather than confirmed malware.

SUSPICIOUS: the skill’s stated purpose is Twitter/X analysis, but all traffic and required auth headers are routed through an unverified third-party gateway on opsat.io rather than official X API infrastructure. There is no overt malware or installer payload, but the undocumented Crab signature requirement, unverifiable ReadX provenance, and lack of publisher-domain linkage make the data flow disproportionate and risky.

SUSPICIOUS. The core capability matches the stated purpose, but the data flow does not: Grok-backed analysis is routed through a custom opsat.io gateway requiring proprietary signature headers instead of xAI's official API. That makes credential handling and data custody opaque, and the skill processes untrusted social content. No local installer or binary execution is present, so this is not confirmed malware, but it is a medium-high risk third-party proxy integration.

SUSPICIOUS. The stated purpose is plausible, but the actual data flow routes all analysis through an undocumented third-party gateway using proprietary authentication rather than official Binance/Bitget endpoints. That mismatch makes the skill's footprint insufficiently transparent and higher risk than its description suggests.