Skills
Navigate
skills/lenialno/roofline/roofline-perf-ru/Gen Agent Trust Hub

roofline-perf-ru

Pass

Audited by Gen Agent Trust Hub on Oct 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute performance analysis tasks using the task (Go Task) runner. Tasks include metrics for monitoring and bench for load testing, targeting a specific local Taskfile.yml located at ${CLAUDE_SKILL_DIR}/../roofline-toolbox/Taskfile.yml.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests performance data from LLM inference servers. Ingestion points: Server metrics from /metrics endpoints and benchmark outputs from task bench (SKILL.md, references/metrics.md). Capability inventory: Local command execution via task (SKILL.md). Boundary markers: None explicitly defined in instructions, though analysis is guided by technical logic tables. Sanitization: None mentioned; the agent is expected to interpret quantitative metrics.
  • [SAFE]: External resources such as documentation for vLLM, SGLang, and roadmap.sh are recognized as well-known and trusted technology services.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 11, 2026, 11:06 AM