graph-sync-yarrrml-review

Graph Sync YARRRML Review

Overview

Use this skill to review or improve YARRRML/RML mappings in graph-sync projects. Focus on correctness, maintainability, runtime compatibility, and semantic output quality.

Use graph-sync-project for broader project edits and runtime configuration.

Review Focus

Check mappings for:

• Relative XPath selectors.
• No duplicate mappings.
• Route/fallback compatibility with the active profile.
• Fallback output policy: catch-all/default mappings should not emit generic WebPage entities unless explicitly intended.
• Template variant expectations for .j2, .liquid, or plain mapping files.
• Morph-KGC compatibility.
• Clear, minimal mapping rules.
• Extraction rules that generalize across pages in the configured source.
• Required vs optional field handling: selective sources should prevent empty triples, while validation should flag missing required fields.
• Canonicalizer-friendly IDs: dynamic subject IRIs should be simple staging IDs, and root entities should expose type, schema:url, and schema:name/schema:headline.
• Postprocessor escalation: keep stable scalar fields in YARRRML, but flag rich text cleanup, nested FAQ/video nodes, API/XHR enrichment, deduplication, validation, or shared logic as postprocessor candidates.
• Avoidance of hardcoded constants from sample pages.
• Explicit IRIs and no blank nodes in emitted entities.
• URL-valued schema properties as plain literals.
• More specific schema.org types instead of generic WebPage when possible.

Process

When reviewing:

• Inspect the relevant mapping file and profile configuration.
• Identify the source pages or representative samples used by the mapping.
• Confirm selectors are grounded in real source structure.
• Check whether XHR/network data offers a more stable source than HTML parsing.
• Review the generated TTL or debug output when available.
• Lead with concrete findings, ordered by severity.
• Include file/line references when possible.

Guardrails

• Do not invent unsupported function syntax.
• Do not use JSON-LD or other structured data markup as an extraction source.
• Do not introduce absolute XPath selectors.
• Do not use a root/page iterator to emit optional or conditionally present values that may become empty literals or malformed IRIs.
• Do not silently downgrade expected authorship or other required creative-work properties; missing required data is a validation issue.
• Do not add hardcoded fallbacks unless the user explicitly authorizes them.
• Escalate when a requirement cannot be implemented cleanly with current mapping capabilities.