The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires administrative privileges to collect detailed per-process GPU and power metrics.\n
Evidence: Use of sudo powermetrics --samplers tasks --show-process-gpu -n 1 -i 1000 in SKILL.md.\n- [EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to install third-party software packages from an external repository.\n
Evidence: Command brew install macmon bottom samply hyperfine in SKILL.md.\n- [COMMAND_EXECUTION]: The skill provides mechanisms to execute or attach to local binaries for performance profiling, which involves executing potentially arbitrary local binaries.\n
Evidence: samply record ./my-program and samply record --pid 12345 in SKILL.md.\n- [PROMPT_INJECTION]: There is an attack surface for indirect prompt injection because the skill ingests data from external command outputs (like ps or top) without sanitization or boundary markers.\n
Ingestion points: Tool outputs from ps, top, and macmon (SKILL.md).\n
Boundary markers: None identified.\n
Capability inventory: Includes sudo command execution and package installation (SKILL.md).\n
Sanitization: No evidence of output filtering or sanitization before processing.

macos-performance-triage