Security audit
Security checks for vulnerabilities and agentic risk
The skill's market-alert purpose is clear, but it asks users to run unshipped local code, configure scheduled delivery, and can send Telegram alerts from broad conversational triggers.
Review this before installing. Only use it with explicit pkedge commands, inspect the referenced Node scripts before running npm install or node cron.js, and avoid launchd boot scheduling unless you are comfortable with persistent Telegram alert delivery using your bot credentials.
The trigger set includes broad natural-language phrases that can activate the skill during ordinary discussion about prediction markets rather than an explicit user request to run it. In this skill, unintended activation is more concerning because it can launch web searches, market analysis, and outbound Telegram delivery workflows tied to scheduled or premium-style actions.
The DEEP DIVE triggers use open-ended phrases like asking what the skill 'thinks' about a topic or whether something is 'mispriced,' which are likely to overlap with normal analytical conversation. Because DEEP DIVE performs multiple searches, inference, and trader-channel delivery, accidental invocation could cause unintended external actions and expose analysis to the wrong workflow.
Phrases like 'whale watch' and 'what are the whales doing' are highly generic and can be used in ordinary conversation unrelated to this skill. In this context, accidental activation would trigger scanning of top trader wallets and generation of trader-only intelligence, increasing the chance of unintended data collection, rate-limit consumption, or message delivery.
The INSIDER WATCH triggers are especially risky because they include vague suspicious-activity language such as 'suspicious wallets' and 'any insider activity,' which can match many unrelated user prompts. This mode performs broad monitoring of wallets and trades, labels patterns as insider-like, and can emit real-time alerts, so accidental execution has higher reputational and operational impact than a read-only query.
No suspicious patterns detected.
