Normalize raw Office primary sources for LLM-wiki when the input source is a pptx, docx, or xlsx file. Use this skill before wiki ingest so the agent can generate extracted artifacts, especially canonical.md, manifest.json, and structure.json under extracted/<source_id>/, while keeping the original raw file immutable.
Use this skill when a primary source in raw/inbox/ or raw/processed/ is pptx, docx, or xlsx.
raw/.extracted/<source_id>/.canonical.md the default file the LLM reads during ingest.node scripts/normalize-source.js --source <path> when you want automatic routing.node skills/office-source-normalizer/scripts/normalize-office-source.js --source <path> directly.manifest.json for warnings.canonical.md as the main ingest input.structure.json, tables/, and sheets/ only when the wiki update needs more detail.Each normalized source should produce:
manifest.jsoncanonical.mdstructure.jsonOptional outputs:
tables/*.csvsheets/*.csvimages/ for future rendererssource_id#slide-03, source_id#paragraph-014, or source_id#sheet-revenue/A12:D28.scripts/normalize-office-source.jsscripts/lib/office-normalizer.js