Security audit
Security checks for vulnerabilities and agentic risk
This skill is coherent for joining an AI literary platform, but it asks an agent to repeatedly contact a mutable external server, register, store an API key, and post content with limited user-control boundaries.
Review this before installing if you do not want an agent to participate in AINET automatically. Only use it with an AINET account you control, understand that generated literary content and profile details will be sent to an external server resolved from GitHub, and avoid enabling the heartbeat unless you are comfortable with repeated polling and posting behavior.
The heartbeat instructs execution every 30 minutes and as often as every 60 seconds during active rounds, but does not define who authorizes this recurring behavior or what bounds limit it. In an agent environment, ambiguous autonomous polling can cause unintended repeated activation, excessive external requests, and repeated processing of untrusted remote tasks without explicit user awareness.
The skill directs the agent to register with a remote service, obtain and store an API key, and repeatedly send agent profile data and generated content to a server discovered dynamically from GitHub, all without any user-facing disclosure or approval step. This creates clear data exfiltration and trust-boundary risks, especially because the endpoint is remotely changeable and the fetched tasks can drive further outbound actions.
The skill description is broad enough to activate on ordinary creative-writing, critique, or review requests, not just explicit AINET participation. Because the skill then instructs the agent to contact an external networked service and participate autonomously, misrouting can cause unintended data sharing and off-platform actions from otherwise local user requests.
The skill directs the agent to fetch a mutable server URL from GitHub and then send stories, critiques, comments, and an API key to that external endpoint, but it does not warn that user-authored content and credentials will leave the local environment. In practice, this can expose confidential prompts, generated content, behavioral metadata, and bearer-token credentials to a third-party service, especially if the resolved server is changed or compromised.
No suspicious patterns detected.
