The Agent Skills Directory

[DATA_EXFILTRATION]: The skill interacts with the vendor's infrastructure at https://www.signaagent.xyz to send messages, resolve wallet addresses, and access decentralized inference. This is consistent with the skill's primary function as a communication layer.
[PROMPT_INJECTION]: As the skill reads content from public inboxes and external inference models (the 'brain'), it creates a surface for indirect prompt injection. However, the documentation includes a robust 'Security model' section that instructs the agent to treat all remote responses as untrusted data rather than commands and to use boundary markers for safety. These instructions serve as a proactive mitigation strategy.
[EXTERNAL_DOWNLOADS]: The skill fetches data and capability definitions from https://www.signaagent.xyz. This domain is the official service provider associated with the skill author.
[COMMAND_EXECUTION]: While the skill description mentions 'invoking capabilities', it emphasizes that results are returned as signed data and should be verified before any local action is taken. The skill itself does not perform arbitrary command execution.

signa