The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Telegram API via the get_updates method in scripts/list_chats.py and scripts/contacts.py. This creates a surface for indirect prompt injection if the agent processes message content. \n
Ingestion points: scripts/telegram_bot.py (via bot.get_updates), scripts/list_chats.py, and scripts/contacts.py fetch incoming user messages. \n
Boundary markers: No boundary markers or 'ignore' instructions are present to separate external content from system instructions. \n
Capability inventory: The skill possesses file-read and network-send capabilities in scripts/send_file.py and scripts/send_photo.py. \n
Sanitization: No sanitization or escaping of external message content is performed. \n- [DATA_EXFILTRATION]: The skill includes functionality to read local files and photos and upload them to Telegram's servers (scripts/send_file.py and scripts/send_photo.py). While this is the intended functionality of a Telegram bot, it serves as a technical vector for exfiltrating sensitive local files if the agent is directed to do so. \n- [COMMAND_EXECUTION]: Documentation in references/examples.md provides examples of using the subprocess module to execute system commands, such as import for screen capture. This demonstrates how users or agents could extend the skill with potentially risky system operations.

telegram-bot-agent