The Agent Skills Directory

[SAFE]: The skill establishes restrictive operational boundaries by limiting file modifications to project-specific paths and designated temporary directories (Section 3.2), and requiring user consent for changes to environment configurations (Section 3.1).
[SAFE]: It minimizes the attack surface by discouraging general-purpose command execution via bash (Section 2.3) and explicitly forbidding raw network requests like curl for fetching documentation (Section 5.2).
[SAFE]: Section 4.3 enforces integrity in the development process by prohibiting the use of comments or flags to suppress linter warnings, the use of unsafe type casting to bypass compiler checks, or the modification of dependencies without permission.

standard