The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests Umbraco schema data from an external CMS via MCP tools (e.g., get-document-type-by-id). This data, such as property descriptions or names, is used to generate requirements documentation and subsequently apply changes back to the site. A compromised site could contain malicious instructions designed to influence the agent's logic.
Ingestion points: references/inspect-existing-schema.md identifies tools like get-document-type-by-id, get-all-data-types, and get-template as sources of external schema data.
Boundary markers: The skill uses strict markdown templates (found in the assets/ folder) and a specific header structure to delimit data.
Capability inventory: The skill possesses the capability to write files to the local filesystem and execute Umbraco schema modifications using MCP tools like create-document-type and update-document-type (as documented in references/apply-via-mcp.md).
Sanitization: The scripts/lint-requirements.mjs utility performs validation on the generated documentation, checking for leftover placeholders and characters that are unsafe for the Umbraco MCP (e.g., ?, &, ../).
[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Node.js script (scripts/lint-requirements.mjs) to validate the generated markdown requirements.
Evidence: SKILL.md and references/change-workflow.md contain instructions to run node scripts/lint-requirements.mjs <requirements-root>. The script performs static analysis on local files and does not involve network activity or dynamic code evaluation.

umbraco-content-requirements-documentation