The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the agent-discordbot CLI tool (distributed via the agent-messenger Node.js package) to execute various Discord operations. The instructions provide numerous examples of using the Bash tool to run commands for authentication, message sending, and data retrieval.
[DATA_EXFILTRATION]: The skill includes functionality for uploading local files to Discord channels via the agent-discordbot file upload command. While this represents a data transmission capability to an external service, it is a core feature of the intended tool functionality.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it is designed to read and process message content from Discord channels (via message list and snapshot commands). This is documented as a inherent risk factor for agents processing external communication.
Ingestion points: Channel messages and server metadata retrieved via message list, message get, and snapshot commands (SKILL.md, references/common-patterns.md).
Boundary markers: The instructions advise the agent to verify memorized data and provide guidance on what information to store in MEMORY.md, but no specific delimiters are enforced in the processing scripts.
Capability inventory: The skill allows for writing messages (message send), uploading files (file upload), and creating threads (thread create).
Sanitization: No explicit sanitization or filtering of Discord message content is performed within the provided shell templates; the agent is expected to handle the content as untrusted input.

agent-discordbot