The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands for installation and troubleshooting. This includes file copying (cp), directory creation (mkdir), and log analysis using common utilities (grep, tail). These operations are typical for mod installation and maintenance.
[PRIVILEGE_ESCALATION]: The troubleshooting section includes a command to modify system firewall rules using sudo ufw allow 7777/udp. While this uses elevated privileges, it is a standard and necessary step for hosting peer-to-peer multiplayer game sessions.
[EXTERNAL_DOWNLOADS]: The mod setup involves downloading BepInEx from its official GitHub repository. GitHub is a well-known service, and referencing the official releases of a widely used modding framework is considered a safe practice.
[INDIRECT_PROMPT_INJECTION]: The skill implements an AI-powered narration feature that transmits player-provided context to external LLM APIs (OpenAI and Claude). This creates a vulnerability surface for indirect prompt injection.
Ingestion points: User input provided via the /api_narrate console command as documented in SKILL.md.
Boundary markers: Absent; the provided C# implementation for the NarrativeEngine interpolates the context directly into the prompt message.
Capability inventory: The mod possesses network capabilities via HttpClient to communicate with external API endpoints.
Sanitization: The code examples do not demonstrate sanitization or filtering of the user-provided context before it is sent to the LLM.

subnautica-ii-coop-mod