roofline-capacity-ru
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill automates LLM deployment tasks by executing shell commands via the
taskrunner, referencing an external Taskfile at${CLAUDE_SKILL_DIR}/../roofline-toolbox/Taskfile.yml. This creates a dependency on code logic maintained outside of the skill's own package.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests external configuration data, creating a potential attack surface.\n - Ingestion points: Deployment variables are read from
inference.envand environment-specific files in theprofiles/directory.\n - Boundary markers: No delimiters or specific safety instructions are used to separate configuration data from the agent's instructions.\n
- Capability inventory: The agent can execute system commands including benchmarking, GPU monitoring, and service management tools via the
taskrunner.\n - Sanitization: The skill does not perform validation or escaping of environment variables before using them in shell command arguments.\n- [EXTERNAL_DOWNLOADS]: The skill provides links to technical documentation and educational resources from
roadmap.shandvllm.aito guide users through model sizing, cost estimation, and deployment strategies. These are well-known technical resources.
Audit Metadata