The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages GitHub accounts by executing the git and gh command-line tools. The implementation in scripts/gh_switch.py uses subprocess.run with list-based arguments, ensuring that user-provided values like target repositories or login names are not interpreted by a shell, which mitigates command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by reading project files and repository metadata. Ingestion points include the .env/.env configuration file and the output of git remote -v. Boundary markers are effectively implemented through programmatic parsing and strict input validation rather than direct prompt interpolation. Capability inventory includes reading local files and executing CLI tools. Sanitization is performed using a restrictive regular expression for GitHub logins, which prevents the processing of malicious payloads embedded in the configuration file.
[DYNAMIC_EXECUTION]: The skill's test suite (scripts/test_gh_switch.py) dynamically generates and executes a mock gh utility. This is a standard testing practice using a static template to simulate CLI responses and does not involve executing untrusted code or remote payloads.

gh-switch