The Agent Skills Directory

Remote Code Execution (CRITICAL): Automated scan confirmed a piped remote execution pattern using curl -L https://raw.githubusercontent.com/akuity/kargo/main/hack/quickstart/install.sh | sh. The source organization 'akuity' is not in the trusted list, making this a confirmed critical risk.
Data Exposure (HIGH): The expressions engine exposes a secret() function in references/expressions.md that allows reading sensitive Kubernetes Secrets (e.g., ${{ secret("my-secret").apiKey }}), facilitating credential exposure.
Command Execution (MEDIUM): The documentation includes unsafeQuote() and regex matches capabilities which can be misused to construct malicious payloads for downstream command execution.
Indirect Prompt Injection (LOW): The skill lacks sanitization for untrusted data entry points such as vars and ctx.meta which are interpolated into expressions. Capability inventory includes Kubernetes secret access and HTTP operations. No boundary markers or sanitization logic are documented.

kargo-skill