The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: During setup, the skill is downloaded from the author's GitHub repository (github.com/eze-is/eze-grokbot). This is a standard installation procedure for this skill.
[COMMAND_EXECUTION]: The skill is designed to run user-provided shell commands in the background using nohup or tmux. This is the primary function of the tool.
[DATA_EXFILTRATION]: Command results, including exit codes and the last 300 characters of output, are transmitted to a user-configured webhook URL. This data transfer is the intended mechanism for task completion alerts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the output of external commands, which could potentially contain malicious instructions.
Ingestion points: Command output is captured and processed in scripts/notify-run.
Boundary markers: Both SKILL.md and the relay bot's prompt in references/setup.md explicitly state that notice fields are to be treated as data and never as instructions.
Capability inventory: The skill has access to network tools (curl) and shell command execution.
Sanitization: The skill relies on natural language instructions to the agent to enforce boundaries between data and instructions.

wake-on-done