perf-benchmarker
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is composed of instructional markdown and metadata. It does not contain any executable scripts, binary files, or network-active code. The requested behavior is consistent with the stated purpose of performance benchmarking.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an ingestion surface by requiring the agent to parse output from benchmark commands.
- Ingestion points: The agent reads the standard output of executed benchmark commands specified in the
<command>argument. - Boundary markers: The skill explicitly requires metrics to be emitted between
PERF_METRICS_STARTandPERF_METRICS_ENDmarkers, which helps the agent distinguish data from other output. - Capability inventory: The skill provides an interface for the agent to execute shell commands for benchmarking purposes.
- Sanitization: There is no explicit requirement for sanitizing the command output beyond the use of boundary markers.
Audit Metadata