The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands such as open {project-name}-research-dashboard.html. Because the {project-name} variable is derived from project context or user input, there is a risk of command injection if the input contains shell metacharacters (e.g., ; rm -rf /).
[DATA_EXFILTRATION]: The generated HTML summary files include inline JavaScript that utilizes navigator.clipboard.writeText. This allows for the programmatic transfer of research data and offer details to the system clipboard, which serves as a mechanism for moving data out of the agent's immediate execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill features a high data-ingestion surface, including automated web searches for competitor reviews and forum threads, as well as the processing of user-provided research documents. This content is then used to construct personas and adversarial agents without explicit sanitization or instructions to ignore embedded commands, creating a surface for indirect prompt injection from malicious external sources.

grandslam-offer