The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the presence of a local binary named dcg. It includes a validation script (scripts/validate-dcg.sh) that executes subcommands such as dcg doctor and dcg test to verify the tool's installation and rule efficacy.
[EXTERNAL_DOWNLOADS]: The skill's documentation and help systems reference an update mechanism (dcg update) that downloads components from the official repository at https://github.com/anthropics/destructive-command-guard. This is a standard update procedure for the tool and targets a trusted organization.
[SAFE]: Static analysis flags regarding destructive system commands (e.g., rm -rf /) are identified as false positives in this context. These patterns appear in documentation and test scripts as examples of what the guard tool is designed to block or as validation cases to ensure the security enforcement is functioning correctly.

dcg