The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is susceptible to shell command injection as it interpolates user-provided profile names directly into a shell execution string: gsd-sdk query config-set-model-profile <profile-name> --raw. An attacker can provide malicious input (e.g., budget; malicious_command) to execute arbitrary code on the host system because the input from {{GSD_ARGS}} is not validated or escaped.- [DATA_EXFILTRATION]: The skill discloses the author's local directory structure and account name by including hardcoded absolute file paths in the <execution_context>, such as /Users/patrykattc/work/git/rag-architecture-patterns/.cursor/get-shit-done/workflows/settings.md. This information exposure reveals internal environment details.- [EXTERNAL_DOWNLOADS]: The skill directs users to install an external package from a public registry via npm i -g get-shit-done. While this represents the vendor's own tooling, it involves the retrieval and installation of remote code.

gsd-config