The Agent Skills Directory

[PROMPT_INJECTION]: The instructions implement behavioral steering to override the agent's default tone with a skeptical persona. The intent is to define a specific feedback style for business critiques rather than bypassing core safety guidelines.
[PROMPT_INJECTION]: The skill processes untrusted external content, such as URLs and user-provided post drafts. It lacks explicit instructions for the agent to use boundary markers or delimiters, which creates a potential surface for indirect prompt injection where instructions could be hidden in the analyzed text.

human-like