The Agent Skills Directory

[COMMAND_EXECUTION] (LOW): The skill provides helper functions (e.g., sh.Run, exec.Command) and examples for executing shell commands such as go build, go mod download, and git diff. This is the primary intended function for a build automation skill.
[EXTERNAL_DOWNLOADS] (LOW): The skill references several external Go modules and tools (e.g., github.com/magefile/mage, golangci-lint, buf). While these originate from non-whitelisted GitHub repositories, they are well-known, industry-standard development tools.
[PROMPT_INJECTION] (LOW): The skill introduces a surface for indirect prompt injection by defining how an agent should read and interpret user-provided Magefiles. A malicious file could potentially contain embedded instructions to exploit the command execution capabilities provided by the skill.
Ingestion points: The agent is instructed to read and understand magefile.go or project-specific build configurations provided by users.
Boundary markers: Absent; no specific delimiters or protective warnings are suggested in the templates to isolate user data from instructions.
Capability inventory: The skill explicitly provides tools for arbitrary system command execution and file system manipulation.
Sanitization: No input validation or sanitization of command-line arguments is performed in the provided helper functions.

way-magefile