The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to discover and execute shell commands for building, testing, and linting software projects. It prioritizes project-defined workflows (e.g., package.json scripts) over generic defaults, which is standard behavior for development automation tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository, including package.json, CI configurations, and documentation. While this creates a theoretical surface for indirect prompt injection, the risk is inherent to the skill's primary function of repository analysis. The skill includes specific guardrails against suppressing errors or bypassing type safety, which serves as a mitigation against accidental or malicious degradation of code quality.

kernel-dev-build