The Agent Skills Directory

[SAFE]: The skill does not exhibit any malicious patterns such as prompt injection, obfuscation, or unauthorized data exfiltration. All external resources and domains (bankr.bot) are associated with the vendor, and no hardcoded credentials or malicious scripts were detected.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages sensitive data like API keys and private wallet information. It explicitly instructs users and agents to follow security best practices, such as storing keys in environment variables, using .gitignore for configuration directories, and setting restrictive file permissions (chmod 600). These are safe and standard management practices.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill recommends the installation of the vendor-scoped CLI tool '@bankr/cli' from the public NPM registry. This is the intended distribution method for the platform's tooling and does not constitute a security risk within the vendor's own skill context.
[INDIRECT_PROMPT_INJECTION]: As a tool that processes natural language and interacts with external web content (web browsing tool) and file storage, the skill possesses an attack surface for indirect prompt injection. However, the documentation highlights safety guardrails, including sanitization of token addresses and verification of contract signatures, which mitigate these risks.
[COMMAND_EXECUTION]: The skill uses shell commands for wallet management and LLM configuration (e.g., 'bankr login', 'bankr llm setup'). These commands are scoped to the vendor's CLI and are used for legitimate setup and operation tasks.

bankr