The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Telegram chats and channels, which are attacker-controlled sources. This creates a potential surface for indirect prompt injection if the agent processes the retrieved content.
Ingestion points: Multiple scripts in SKILL.md (Step 4, 5, 6, 8, 9) fetch messages and metadata via client.iter_messages and client.iter_dialogs.
Boundary markers: No delimiters or "ignore instructions" warnings are used to wrap the content extracted from Telegram messages.
Capability inventory: The skill has capabilities to write files to the local system (download_media, json.dump, f.write) and perform network operations via the Telethon client.
Sanitization: The skill does not sanitize or filter the content of messages before processing or saving them to the filesystem.
[EXTERNAL_DOWNLOADS]: The skill requires installation of third-party Python packages.
Evidence: The instructions prompt the user to run pip install telethon pillow cryptg. These are well-known and reputable libraries for Telegram interaction and image processing.

telegram-export