The Agent Skills Directory

[SAFE]: The skill serves as a scaffolding framework for developing other skills. It includes comprehensive instructions that explicitly forbid the inclusion of secrets, destructive commands, or hidden network automation in any generated content.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and specification sites for agent development, including Claude's platform documentation, Gemini CLI guides, and the Agent Skills specification. These references are for informational purposes and target well-known, trusted technology services.
[REMOTE_CODE_EXECUTION]: The skill explicitly mandates that any scripts generated for new skills must never use remote-code execution patterns or download-and-run behaviors, mitigating supply chain risks in the resulting output.
[COMMAND_EXECUTION]: While the skill involves file system operations (creating and updating files under a 'skills/' directory), these are restricted to the intended scope of skill authoring and include clear approval requirements for any risky actions.

fusion-skill-authoring