The Agent Skills Directory

[Input Scope Validation]: The skill includes a pre-validation step (Step 0) that requires an explicit list of test methods or files to be provided. This design prevents unintended or broad-scope audits of a workspace, ensuring the agent only processes data within a user-defined scope.
[Analysis-Only Context]: Although the skill reads and processes untrusted source code as part of its grading workflow, its capabilities are limited to text processing and pattern matching. It does not involve code execution, subprocess spawning, or network interactions, which minimizes the risk associated with processing external code.
[Dependency Management]: The skill references other modular components (such as test-analysis-extensions) but does not perform external downloads or install third-party packages at runtime. All operations are self-contained within the expected agent environment.

grade-tests