The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the agent-telegram CLI through the Bash tool to perform operations such as authentication, chat listing, and message sending.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes the agent-messenger package from the npm registry using the bunx command at runtime.
[CREDENTIALS_UNSAFE]: The skill manages sensitive Telegram authentication data, including phone numbers, verification codes (used for both credential provisioning and account login), and 2FA passwords. It stores persistent session and state data in the ~/.config/agent-messenger/telegram/ directory.
[PROMPT_INJECTION]: The skill reads external Telegram messages, which represents an attack surface for indirect prompt injection where malicious instructions could be embedded in messages.
Ingestion points: Telegram messages are retrieved using the agent-telegram message list command in SKILL.md.
Boundary markers: No delimiters or specific instructions to ignore embedded commands are defined for the retrieved message content.
Capability inventory: The agent can execute various agent-telegram commands to send messages, search chats, and modify authentication states across ALL scripts.
Sanitization: No validation, filtering, or escaping of the external message content is mentioned in the skill instructions.

agent-telegram