list-telegram-chats
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the agent executing a Python script (
list_chats.py) via the system shell to aggregate data from multiple sources. - [INDIRECT_PROMPT_INJECTION]: The skill processes metadata (such as group titles and user names) retrieved from the Telegram API and local interaction session files. This represents an indirect prompt injection surface if an external actor provides malicious instructions within their Telegram profile or group metadata.
- Ingestion points:
~/.nanobot/workspace/sessions/telegram_*.jsonl, OpenClawsessions.json, and responses from theapi.telegram.org/bot/getChatendpoint. - Boundary markers: Not present in the instruction set; the agent receives the raw JSON output.
- Capability inventory: The agent has shell access (
Bash) to execute the helper script. - Sanitization: The script parses JSON but does not filter or sanitize string content (e.g., chat titles) before returning it to the agent.
- [SAFE]: The network operations are directed to well-known services (Telegram API) or environment-defined backend URLs (
YOUAI_API_URL) required for the skill's primary purpose. Access to local configuration files (~/.nanobot/config.json) is used appropriately to retrieve the necessary Telegram bot token.
Audit Metadata