diff --git a/phases/phase-1-scaffolding.md b/phases/phase-1-scaffolding.md
new file mode 100644
index 0000000..187a479
--- /dev/null
+++ b/phases/phase-1-scaffolding.md
@@ -0,0 +1,96 @@
+# Phase 1: Scaffolding
+
+**Executed:** 2026-04-14T14:45Z UTC
+
+---
+
+## Backups (taken before changes)
+
+| Item | Location | MD5 Hash |
+|------|----------|----------|
+| recon.db | CT 130: `/tmp/recon.db.phase1.20260414.bak` | `69d94a2c21686871c8c6863903710e3f` |
+| config.yaml | CT 130: `/tmp/config.yaml.phase1.20260414.bak` | `6d70ed572dfb2e704abca3850ae33797` |
+
+DB hash matches Phase 0 backup — no changes occurred between phases.
+
+---
+
+## What Changed
+
+### 1. Filesystem: New directory tree
+
+Created under `/opt/recon/data/`:
+
+```
+acquired/
+  README.md
+  pdf/.keep
+  stream/.keep
+  html/.keep
+processing/
+  README.md
+```
+
+All owned by `zvx:zvx`, matching the existing data directory.
+
+### 2. Config: Three edits to `/opt/recon/config.yaml`
+
+**a) `new_pipeline.enabled` set to `false`**
+
+The Stream B library pipeline (watchdog-driven file intake from `_acquired/` and `_ingest/`) is disabled. This prevents the old pipeline from processing files while we build the replacement.
+
+**b) `crawler.sites` set to `[]`**
+
+All 44 crawl target site definitions commented out and preserved as historical reference. The crawler scheduler will find zero sites and do nothing if started.
+
+**c) New `pipeline:` section added at end of file**
+
+```yaml
+pipeline:
+  acquired_root: /opt/recon/data/acquired
+  processing_root: /opt/recon/data/processing
+  dispatch:
+    pdf: pdf_processor
+    stream: transcript_processor
+    html: html_processor
+  mtime_stability_seconds: 10
+```
+
+Scaffolding only — no code reads this section yet. Processors do not exist.
+
+**Config diff stats:** 284 lines removed, 302 lines added (bulk is the 44 sites being commented/uncommented).
+
+### 3. Schema: `text_dir` column added to `documents` table
+
+```sql
+ALTER TABLE documents ADD COLUMN text_dir TEXT;
+```
+
+All 29,812 existing rows have `text_dir = NULL`. This column will hold the path to each document's extracted text directory, replacing the convention-based `data/text/{hash}/` lookup.
+
+---
+
+## What Did Not Change
+
+- **No code modified:** `recon.py`, `lib/`, `scripts/`, templates, static assets — all untouched
+- **No data modified:** catalogue and documents row counts remain 29,812 each
+- **No service state changed:** Both `recon.service` and `recon-watchdog.service` remain inactive (both still `enabled` — will auto-start on reboot)
+- **No Qdrant changes:** Collection `recon_knowledge_hybrid` untouched (2,320,695 points)
+- **No file moves or deletions:** Existing `data/text/`, `data/concepts/`, NFS mounts all untouched
+
+---
+
+## Verification (post-change)
+
+| Check | Result |
+|-------|--------|
+| recon.service | inactive |
+| recon-watchdog.service | inactive |
+| catalogue rows | 29,812 |
+| documents rows | 29,812 |
+| text_dir NULL count | 29,812 (all rows) |
+| new_pipeline.enabled | `false` |
+| crawler.sites | `[]` |
+| pipeline.acquired_root | `/opt/recon/data/acquired` |
+| New directories exist | all 5 confirmed, zvx:zvx |
+| YAML validates | yes |