The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user requests to modify site schemas, which is an inherent attack surface for indirect prompt injection. However, the risk is effectively mitigated by the skill's mandatory human-in-the-loop workflow.
Ingestion points: User instructions and existing site schema retrieved via MCP tools in workflow.md.
Boundary markers: The skill uses a requirements documentation format and an explicit 'Stop for approval' gate as defined in workflow.md step 5.
Capability inventory: Employs MCP tools such as update-document-type, move-document-type, and create-document-type-folder to modify the Umbraco environment.
Sanitization: Relies on a 'Read, modify, write back' pattern where the agent must compare the proposed requirements doc against a fresh read of the site schema.
[COMMAND_EXECUTION]: The skill uses the Umbraco Developer MCP to perform schema updates. These operations are standard within the Umbraco ecosystem and are protected by the requirement for user approval before execution, as specified in the Best practices and Steps sections of SKILL.md.

umbraco-change-document-type