The Agent Skills Directory

Safe Command Execution: The skill interacts with the local operating system by invoking the git, gh, and az command-line tools. These interactions are performed using the System.Diagnostics.Process API with explicit argument lists rather than shell-mediated execution. This context-aware approach effectively prevents command injection vulnerabilities that could arise from processing external or untrusted data. \n- Validated External Access: The skill fetches necessary build and package metadata from established and trusted services, specifically GitHub and Azure DevOps Artifacts. These network operations are strictly limited to these domains and are essential for retrieving the release signals required for readiness assessment. \n- Proactive Data Sanitization: A comprehensive sanitization module (PublicReportSanitizer.ps1) is integrated to identify and redact internal tool names, private URLs, and sensitive identifiers. It uses advanced techniques to detect varied encodings—such as homoglyphs, HTML entities, and URL encoding—ensuring that private release details are sanitized before they are published in public tracker issues. \n- Mitigation of Indirect Prompt Injection: The skill processes data from external sources, such as pull request titles and issue descriptions, which represents a potential indirect prompt injection surface. The skill mitigates this by escaping markdown-significant characters and pipe delimiters in its reports and by using explicit boundary markers () to prevent automated content from being manipulated by malicious data. \n- Secure Credential Management: The skill implementation and documentation prioritize the security of sensitive tokens. It leverages standard environment variables for managing Azure DevOps PATs and includes explicit warnings and code logic to prevent credentials from being exposed in command-line arguments, reports, or logs.

release-readiness