The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the bale-mcp package using the uvx tool. This package is downloaded from an external registry (PyPI) and provides the core logic for the agent's interaction with the Bale API.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to read and respond to messages from external users on the Bale messenger, which introduces an attack surface for indirect prompt injection. Malicious messages could potentially influence the agent's behavior.
Ingestion points: Untrusted content is ingested via tools such as bot_get_new_messages, bot_read_chat, bot_search_messages, and bot_wait_for_reply (SKILL.md).
Boundary markers: There are no explicit instructions in the skill to use delimiters or ignore potentially malicious commands embedded within chat messages.
Capability inventory: The agent possesses significant capabilities, including sending messages (bot_send_message), uploading files (bot_send_file), modifying content (bot_edit_message), and downloading files from the Bale servers to the local environment (bot_download_file).
Sanitization: No content sanitization or validation logic is defined within the skill to mitigate the risk of processing malicious text or data from external chats.

bale