The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill's behavior matches its stated purpose of assisting with Telegram management.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and summarizes unread messages from external Telegram users. Malicious content within these messages could attempt to override the agent's summarization instructions.
Ingestion points: Messages are fetched using list_messages, get_messages, and search_messages tools in SKILL.md.
Boundary markers: The instructions do not use explicit delimiters (like XML tags) to wrap the external message content when the agent processes it.
Capability inventory: The agent has permissions for file system modification (Write, Edit) to maintain style guides and can interact with the Telegram account via the MCP server.
Sanitization: There is no programmatic sanitization of message content; however, the skill mandates that all outgoing actions be saved as drafts for human review.
[CREDENTIALS_UNSAFE]: The documentation in references/setup.md handles the risk of credential exposure appropriately by instructing the user on secure storage practices (e.g., using .env files and the macOS Keychain) and warning against sharing session strings.

telegram-assistant