The Agent Skills Directory

[SAFE]: The skill consists of instructional templates and documentation for establishing an evaluation harness. Analysis of the content reveals no malicious patterns, prompt injections, or unauthorized data access attempts.
[COMMAND_EXECUTION]: The framework includes examples of using shell commands for deterministic grading, such as npm test, npm run build, and grep. These are standard software development practices for verifying code integrity and do not constitute a security risk in the context of a development-focused evaluation tool.

eval-harness