The Agent Skills Directory

[COMMAND_EXECUTION]: The skill suggests utilizing standard diagnostic and profiling utilities including uptime, nproc, top, pidstat, strace, node, py-spy, and perf. These tools are appropriately recommended for performance monitoring and do not indicate malicious intent.
[INDIRECT_PROMPT_INJECTION]: The skill instructions involve analyzing external content such as measurement scripts and benchmark results, which constitutes an ingestion surface.
Ingestion points: The agent is directed to read measurement scripts and evaluate performance data as described in SKILL.md.
Boundary markers: No specific delimiters or boundary markers for untrusted data are specified in the instructions.
Capability inventory: The skill leverages standard diagnostic tools (uptime, top, strace, etc.) to gather and report performance metrics.
Sanitization: The instructions do not mention explicit sanitization or validation of the ingested scripts or performance data.

benchmark-checklist