The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill directs the agent to install the tdl dependency using the command curl -sSL https://docs.iyear.me/tdl/install.sh | sudo bash. This pattern executes a script directly from a remote server without verification, which is a high-risk behavior.
[COMMAND_EXECUTION]: The installation process uses sudo, granting administrative privileges to a third-party script. This represents a privilege escalation risk where the script could modify system-level files or install persistent backdoors.
[EXTERNAL_DOWNLOADS]: The skill fetches software and configurations from docs.iyear.me and github.com/iyear/tdl. Since iyear is not an identified trusted organization or well-known cloud service, these external dependencies are considered unverifiable.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it ingests and processes untrusted data from external Telegram channels.
Ingestion points: Telegram message content exported to /tmp/tdl-export.json and read via cat in SKILL.md (Step 4).
Boundary markers: Absent; there are no instructions to the agent to treat the imported Telegram content as untrusted data or to ignore instructions embedded within the messages.
Capability inventory: The skill has the ability to execute shell commands (tdl), read files, and initiate network requests (via the tdl tool).
Sanitization: There is no evidence of sanitization or filtering of the Telegram content before it is processed by the agent.

telegram