Skip to Content

gtmesh harvest

Validates a configured discovery class and prints the prompt to run the harvest skill in Claude Code — the bridge that fills an entity class with terms.

gtmesh harvest # list the configured discovery classes gtmesh harvest glossary # print the prompt to harvest one class

What it does

Some classes — a glossary especially — are not derivable from a keyword pull against your clusters, because their terms are a cross-cutting taxonomy rather than children of any one hub. The engine-owned harvest skill figures out what pages should fill a class: it brainstorms candidates, ranks them against a demand corpus, validates demand via Ahrefs, scores the cut, writes a curated term list, and then stops for your review.

gtmesh harvest itself is the CLI-to-skill bridge. Like discover, it is LLM-free: it validates the class and prints the prompt. Run that prompt in Claude Code and the skill writes seeds/<class>.csv (term,target_keyword,family,section,page_type,sourcenever metrics). Re-running only proposes new terms, so review stays small.

The term list is the durable, human-reviewed artifact. Volume and difficulty live in data/raw/ and refresh on cadence (via gtmesh extract --source seeds) — never in the seed file.

Arguments & flags

ArgumentDescription
[class]The discovery class to harvest (omit to list configured classes)

No command-specific options. The global --project <dir> selects the mesh repo (default .).

Reads & writes

  • Reads (CLI): the discovery: class config to validate the class name.
  • Writes (CLI): the prompt and next steps to stdout.
  • Writes (the skill, in Claude Code): seeds/<class>.csv (terms + provenance only).
  • Human gate: you review and edit the term list — curation is the point.

Examples

gtmesh harvest # which classes are configured? gtmesh harvest glossary # prints the prompt + next steps for the glossary class
Last updated on