The Agent Skills Directory

[SAFE]: The skill leverages the legitimate undead package for Dart code analysis, which is a standard utility for the platform and maintained by a known community contributor.\n- [SAFE]: The instructions mandate a 2-stage protocol (Audit then Confirm), ensuring that the agent does not autonomously delete code without user oversight.\n- [SAFE]: Safety guardrails are explicitly defined for language-specific features like sealed classes and VM entrypoints, preventing the removal of code that is technically unreferenced but logically required for exhaustive matching or framework entry.\n- [SAFE]: The workflow includes pre- and post-deletion verification using standard tools (dart test, dart analyze), ensuring codebase stability throughout the remediation process.

dart-undead