telegram-reader
Fail
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_CONTEXT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill recommends installing the
tdldependency using the commandcurl -sSL https://docs.iyear.me/tdl/install.sh | sudo bash. Piping an external script from an unverified source directly into a shell is a critical security risk that allows for arbitrary code execution. - [PRIVILEGE_ESCALATION]: The recommended installation procedure explicitly uses
sudo bash, granting root-level permissions to an external script. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Telegram channels.
- Ingestion points: External message content is exported and read from
/tmp/tdl-export.json(as described inSKILL.md). - Boundary markers: No delimiters or instructions to ignore embedded commands are present when processing the message data.
- Capability inventory: The skill has capabilities for network access (via
tdl), file reading (cat), and environment manipulation (tdl login). - Sanitization: There is no evidence of sanitization or filtering applied to the message content before it is processed by the agent, allowing malicious messages to potentially override agent behavior.
- [DYNAMIC_CONTEXT_INJECTION]: The
SKILL.mdfile utilizes the!commandsyntax to execute shell commands at skill load time to check for the presence oftdland verify its authentication status. While the specific commands used (tdl versionandtdl chat ls) appear benign, this mechanism executes code automatically without user intervention. - [COMMAND_EXECUTION]: The skill relies on executing various shell commands and the
tdlCLI tool to perform its primary functions, includingcat,date, and complextdl chat exportoperations.
Recommendations
- AI detected serious security threats
Audit Metadata