The Agent Skills Directory

[COMMAND_EXECUTION]: The skill launches a local Python script (scripts/watch_job.py) in the background to monitor system state. It also instructs the agent on the use of standard system tools like ps, lsof, and kill for diagnostic and recovery purposes. These commands are integral to the skill's monitoring functionality and are guarded by explicit instructions to seek user confirmation before execution.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface through the monitoring of external log and output files for completion markers.
Ingestion points: scripts/watch_job.py monitors file growth and log content for specific terminal strings (e.g., "DONE").
Boundary markers: None are applied to the external file content being monitored.
Capability inventory: The skill utilizes run_in_background and provides instructions for process management via kill/pkill.
Sanitization: No sanitization is performed on the data read from external files. However, the risk is categorized as safe because the script primarily reports metadata (status and size) rather than re-injecting raw file content into the agent's prompt context.

job-babysitter