compare-race
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes local command-line interface (CLI) tools, including
compare-race run,compare-race config, andcompare-race record, to orchestrate the benchmarking process and manage race data. - [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and follow external instructions from prompt files (e.g.,
prompts/RACE-STARTER.en.md) and evaluate output stored in local race files. - Ingestion points: External prompt files and generated race result files (
RACE.md, RUN files) are read into the agent's context. - Boundary markers: The skill does not define specific delimiters or instructions to ignore nested commands within these external files.
- Capability inventory: The skill utilizes shell execution, local file system access (read/write), and network access to the vendor's repository.
- Sanitization: No explicit sanitization or filtering of the external prompt content is described.
- [EXTERNAL_DOWNLOADS]: The skill references the official GitHub repository of the vendor (
ellmos-ai) for its source code and configuration examples.
Audit Metadata