The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill ingests untrusted data from Telegram messages and saves it in an 'LLM-friendly' format specifically intended for agent analysis. This creates a high-risk surface for indirect prompt injection.
Ingestion points: Data enters from Telegram API via telegram_sync.py and is written to local Markdown files (e.g., data/{group_id}/messages.md).
Boundary markers: Absent. The described message format lacks delimiters or instructions to ignore embedded commands within message bodies.
Capability inventory: While this skill primarily performs write operations, it is designed to be used with telegram-read. If the agent possesses further capabilities (e.g., shell access or file modification), messages from Telegram could trigger those actions.
Sanitization: No mention of sanitization, filtering, or escaping of message content.
Data Exposure (HIGH): The skill targets highly sensitive personal information, including Direct Messages (DMs), which are synced by default. This exposes private communications to the agent's environment, increasing the risk of data leakage if the agent is subsequently compromised.
Command Execution (LOW): The skill requires the execution of multiple Python scripts (telegram_sync.py, telegram_list.py, telegram_init.py) via the command line. While legitimate for the skill's function, it represents the primary mechanism for the sensitive data access.

telegram-sync