| name | knowledge-harvest |
| description | Use to sweep the CURRENT repo for knowledge references that are not yet catalogued — docs/*.md research files and http links in README/docs — and turn the chosen ones into registry entries by domain. The batch-capture companion to knowledge-registry, parallel to project-scan. Triggers — "cosecha las referencias", "escanea el repo por docs/links", "qué conocimiento no está registrado", "harvest knowledge", "/knowledge harvest". |
| user-invocable | false |
knowledge-harvest
Batch-capture references already scattered in the repo into knowledge/registry.json.
Workflow
- Scan for candidates (read-only, dedups against the registry):
pwsh -File ${CLAUDE_PLUGIN_ROOT}/scripts/Invoke-KnowledgeHarvest.ps1 -Root . -Json
- Present the candidates grouped by kind. For each one the user wants to keep, ask which
domain it belongs to (offer the registry's declared domains;
-NewDomain to extend).
- Add each chosen candidate via the registry engine:
pwsh -File ${CLAUDE_PLUGIN_ROOT}/scripts/Add-KnowledgeRef.ps1 -Root . -Ref <ref> -Domain <Domain> -Title "<title>" -Note "<one line>"
- Show the regenerated table (
Get-KnowledgeInventory.ps1).
When NOT to use
- A single known reference → use
knowledge-registry (/knowledge add) directly.