The Agent Skills Directory

[COMMAND_EXECUTION]: The script check.sh executes several macOS system utilities (sysctl, vm_stat, df, ps) to gather performance and resource metrics. This is required for the skill's primary function and operates with user privileges.
[PROMPT_INJECTION]: The skill processes process names from the system which are then provided to the agent. This creates an indirect prompt injection surface where a maliciously named process could attempt to influence the agent's behavior during report generation.
Ingestion points: The script check.sh reads the ucomm (process accounting name) field from the output of the ps command.
Boundary markers: While the data is passed in JSON format, the instructions for the agent turn in openclaw-cron.json do not include explicit delimiters or safety instructions to disregard prompts embedded within the system data.
Capability inventory: The skill executes a local shell script and provides data to an agent capable of using a messaging tool.
Sanitization: Process names are trimmed but not checked for instruction-like content. The attack surface is mitigated by the use of the ucomm field, which is typically restricted to the executable's short filename rather than full command-line arguments.

system-watchdog