The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands and scripts to perform 'lightweight testing' and verify code fixes. This functionality is the primary purpose of the skill and is performed within an isolated directory (tmp/run_*/) to minimize impact on the host system.\n- [DATA_EXFILTRATION]: The instructions include a procedure for reporting bugs found in the skill itself to a specific GitHub repository (huangwb8/bensz-bugs) using the GitHub CLI (gh). This action is strictly conditional on the user explicitly requesting a bug report, ensuring that no data is sent externally without user consent.\n- [DATA_EXPOSURE]: The skill ingests and analyzes a wide range of project files, including source code, Dockerfiles, and CI/CD configurations. However, all analysis results and intermediate artifacts are stored locally within the project's 'tmp' directory, and there is no evidence of unauthorized data transfer.\n- [REMOTE_CODE_EXECUTION]: The skill uses local Python scripts (create_session.py, verify_session.py) for workflow management. There are no patterns indicating the download or execution of unverified remote code or third-party packages at runtime.

auto-test-code