The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess execution to run various system diagnostic tools including sysctl, system_profiler, ps, df, du, and pmset.
[DATA_EXFILTRATION]: Inspects sensitive system paths and metadata, specifically reading contents or sizes from LaunchAgents, startup items, and the user Downloads directory.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted system data (e.g., process names, file paths) that could contain malicious instructions.
Ingestion points: System command outputs from ps, du, and directory listings in SKILL.md.
Boundary markers: None present to isolate ingested system data from the agent's internal instruction set.
Capability inventory: Extensive execution capabilities for system diagnostic utilities and shell-based data processing.
Sanitization: No sanitization or validation logic is applied to the data retrieved from system commands before it is interpreted by the agent.

mac-analyze