The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided markdown content via the get_page tool to generate HTML files. While ingesting untrusted content is a common attack surface, the skill explicitly mandates a sanitization process. The gbrain publish tool is instructed to strip sensitive internal metadata, citations, and PII (Personal Identifiable Information) before the HTML is created, which mitigates the risk of malicious data influencing downstream processes.
[SAFE]: The skill enforces security best practices for data handling. It defaults to AES-256-GCM encryption with PBKDF2 key derivation for all published files and explicitly instructs the agent to inform the user to share passwords through a different communication channel than the file URL.
[SAFE]: The command execution patterns involve vendor-specific CLI tools (gbrain) for file generation and cloud storage management, which are consistent with the skill's stated purpose and author (praxstack).

publish