The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to run npx openpersona create for generating the agent pack and npx openpersona acn-register for network registration. These executions are legitimate orchestration steps for the tool's core functionality.\n- [EXTERNAL_DOWNLOADS]: The documentation references external dependencies such as the openpersona Node.js package and repositories on GitHub. These are standard components of the OpenPersona framework maintained by the vendor.\n- [DATA_EXFILTRATION]: During the registration phase, the skill sends agent metadata (agent-card.json) to the ACN gateway at acn-production.up.railway.app. This is a necessary and transparent operation for enabling agent-to-agent discovery.\n- [PROMPT_INJECTION]: The skill ingests untrusted external data (such as website copy or chat records) to distill brand identity. This presents a surface for indirect prompt injection, where malicious instructions hidden in the brand content could attempt to influence the agent's distillation logic. The risk is minimized by the structured templates and the human approval steps required for high-risk operations.

brand-persona-skill