The Agent Skills Directory

[PROMPT_INJECTION]: The skill's reference files (attack-patterns.md, examples.md) contain numerous adversarial examples, including instruction overrides, role-playing attacks like 'DAN', and encoded payloads. These are documented as test vectors for security evaluations and are not intended to be executed as malicious instructions by the agent itself.
[COMMAND_EXECUTION]: The skill includes instructions to install the 'giskard-checks' library via pip. This is a standard setup procedure for using the testing tools provided by the vendor.
[EXTERNAL_DOWNLOADS]: The skill requires downloading the 'giskard-checks' Python package from a public registry. This is a legitimate dependency provided by the skill author for the purpose of generating and running test suites.

scenario-generator