memorb-ingestlisted

# Memorb Ingest Skill > LLM Wiki Architecture: **raw (sources) → wiki (memorbs/ structured pages) → schema (CLAUDE.md)**. > Ingest ensures consistency across all three layers when new raw material arrives. ## Orb Layout Decision (apply before any write) When creating a **new** orb in `memorbs/Long-Term/` or `memorbs/HQ/`: - **Has attachments** (images, PDFs, audio, etc.) → create a **bundle orb**: `mkdir {base}/{name}/` then write `{base}/{name}/{name}.md`; place attachment files in the same folder. - **No attachments** → create a **plain orb**: write `{base}/{name}.md` directly. When **updating an existing** orb, resolve its path first (two-step lookup): 1. Check `{base}/{name}.md` — if found, it is a plain orb; update it. 2. If not found, check `{base}/{name}/{name}.md` — if found, it is a bundle orb; update the inner `.md` file. ## Execution Sequence (Strict Order) 1. **Distill into an orb**: - The distilled orb lands in `memorbs/HQ/OrbTrack/` unless it clearly belongs to an existing entity page. - If there is no raw material to keep, it is a plain orb: `{name}.md`. 2. **Keep the raw material inside the orb's bundle** *(only when there is any)*: - Make the orb a bundle — `{name}/{name}.md` — and drop the transcript, clipping, or PDF alongside it in the same folder. Nothing goes to a shared attachment bucket and nothing is written outside `memorbs/`. - **Documents only** (`.md`, `.txt`, `.pdf`, document scans). Audio and video never enter the vault; record