docs(our-gdrive-organizer): add 3 gotchas from Brand in Action audit

- D_intelligence (underscore) — typo, NOT a regex variant. Fix one-off
  via mv; do not widen RENAME_RULES (false-positive risk on legit
  underscore-separator filenames).
- Externally-generated filenames vs OurDigital convention. Default to
  normalize {Client}_/Client_/client_ → CLIENT- inside Active Workspaces;
  drop 14-digit timestamps to YYYYMMDD; preserve only when an external
  system requires the literal name.
- Reference library naming inconsistency — DO NOT bulk-normalize. Mixed
  naming reflects source provenance (Slideshare slugs, vendor whitepapers,
  Korean blog captures). Group by topic into subfolders instead
  (frameworks/, examples/, ko/) when count exceeds ~15 at root.

Patterns library now at 15 gotchas. Validated by live application during
01_Brand in Action audit.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

custom-skills/82-our-gdrive-organizer/shared/patterns/gotchas.md

@@ -254,6 +254,93 @@ reference visualization that has real value AND a meaningful home elsewhere.
 
 ---
 
+### `D_intelligence` (underscore) — typo, NOT a regex variant
+
+**Pattern**: A file uses `D_intelligence` with an underscore separator
+between `D` and `intelligence` instead of a dot. E.g.,
+`D_intelligence_홍아현_프로젝트부속계약서_20260501.gdoc`.
+
+**Why tricky**: It LOOKS like a third regex-escaping variant of the brand
+prefix (alongside `D.intelligence` with dot and `D intelligence` with
+space, both of which ARE valid spellings the parent company uses). But
+underscore is **wrong** — D.intelligence is always written with a dot.
+The underscore form is purely an accidental typo (probably auto-generated
+by a system that sanitized dots to underscores).
+
+**Resolution**:
+1. Treat as a one-off typo fix: rename `D_intelligence` → `D.intelligence`
+   via `mv`. The rest of the filename keeps its underscores (those are
+   legitimate field separators in the file's own naming scheme).
+2. **Do NOT add to `RENAME_RULES`.** Adding a `D_intelligence` regex
+   would cause false positives in any filename that uses underscores as
+   field separators between something ending in `D` and something starting
+   with `intelligence` — unlikely but defensively avoid widening the rules.
+3. Same applies to other one-off accidental variants: fix with `mv`,
+   document the fix here, but don't widen the rules.
+
+---
+
+### Externally-generated filenames vs OurDigital convention
+
+**Pattern**: Files in an active client workspace whose names come from
+external systems and don't follow OurDigital convention. E.g.:
+- `Jamie_Naver_Booking_Products_v1.0.gsheet` (Title_Case_With_Underscores
+  — looks like a Google Sheets name template)
+- `jam_medical_advertising_review_guide_20251119154552.gslides`
+  (lowercase + 14-digit timestamp — looks like an export from a
+  publishing tool)
+- `jamie_monthly_sow_standard.docx` (lowercase + underscores — looks
+  like a Word template name)
+
+**Why tricky**: These names are functional and the user might prefer to
+preserve them as-is to maintain traceability with the originating system.
+But they violate the canonical `{ClientPrefix}-{topic}-{date}.{ext}`
+pattern from `canonical-files.md`, which makes the folder feel
+inconsistent.
+
+**Resolution**:
+- The OurDigital convention wins inside `01_Active Workspaces/{NN_client}/`
+  because the user owns that folder's organization. Normalize:
+  - `Jamie_` → `JAM-`
+  - `jamie_` → `JAM-`
+  - `jam_` → `JAM-`
+  - `_` field separators → `-` (or space inside hyphenated phrases)
+  - Long timestamps `YYYYMMDDhhmmss` → `YYYYMMDD` (drop time portion)
+  - Title_Case_Words → Title Case Words (use spaces inside hyphen-separated phrases)
+- Default to normalizing unless the user says preserve. The originating
+  system can usually re-export with a new name if needed.
+- If the file is truly the system's source-of-truth filename (e.g., a
+  webhook expects an exact filename match), keep it and note WHY in a
+  README inside the same folder.
+
+---
+
+### Reference library naming inconsistency (do NOT normalize)
+
+**Pattern**: A reference folder like `90_Brand Management Reference/`
+contains 20+ image files imported as a batch from various sources, with
+mixed naming conventions (Title_Case, lowercase-hyphens, Korean-spaces,
+auto-generated junk).
+
+**Why tricky**: It's tempting to normalize all the names for consistency.
+But the inconsistency reflects the varied **sources** of the references —
+each filename traces back to its origin (Slideshare slug, vendor
+whitepaper, screenshot of a textbook page, Korean blog).
+
+**Resolution**:
+- **Do not bulk-rename reference filenames** for cosmetic consistency.
+  They have provenance value as-is.
+- DO fix individual filenames that are functionally broken:
+  - Double extensions (`foo.png--example.png` → `foo-example.png`)
+  - Filesystem-illegal characters
+  - Truncated names that lost the meaning
+- DO group by topic into subfolders if the count exceeds ~15 files at the
+  root: `frameworks/`, `examples/`, `ko/` (or whatever fits the content).
+  Subfolders provide structure without touching the legitimate filename
+  diversity.
+
+---
+
 ## Adding new gotchas
 
 When you (Claude) encounter a new ambiguous case during a content-based