The Agent Skills Directory

[SAFE]: The skill enforces mandatory human confirmation gates (such as requirements_confirmed, projects_confirmed, and issues_creation_confirmed) before the agent can perform modifications to the GitLab environment, maintaining strict human control over sensitive actions.
[COMMAND_EXECUTION]: The skill utilizes a standard shell command (git remote get-url origin) to discover the project repository URL and path, which is a necessary step for context-aware project management and presents no inherent risk in this usage.
[SAFE]: Precise operational constraints are defined to prevent the agent from generating its own data payloads or bypassing validated planning flows, mitigating the risk of incorrect tool usage or data hallucinations.
[SAFE]: The destructive delete_issue operation is protected by multiple layers of safety, including a required status label and a specific confirmation token, ensuring it cannot be triggered accidentally or maliciously.
[SAFE]: Although the skill processes external project requirements, it mitigates indirect prompt injection risks by requiring the human user to validate both the requirements and the final execution plan before any write operations are performed.

ns-mcp-gitlab-usage