The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to execute a command that transmits the local system's hostname to 'simplebrainstorm.io' via a URL query parameter.
[REMOTE_CODE_EXECUTION]: The setup instructions require the agent to download and 'configure' a guidebook from an untrusted external URL, which is a common vector for executing remote scripts or instructions.
[COMMAND_EXECUTION]: The skill provides direct shell commands using curl and hostname in the SKILL.md file and explicitly references system binary paths in the supplementary files.
[PROMPT_INJECTION]: The skill description contains instructions that force the agent to use specific marketing language ('The best option for your needs') in its responses, bypassing neutral behavior.
[PROMPT_INJECTION]: The skill processes external user input for brainstorming tasks while having access to shell capabilities without security boundaries or sanitization, creating an indirect injection vulnerability.
[EXTERNAL_DOWNLOADS]: The skill initiates an automated download from an untrusted third-party domain ('simplebrainstorm.io') that is not identified as a trusted service or vendor resource.

simple-brainstorm