update marker context with DD S2 hybrid approach context

marker-context.md

@@ -180,3 +180,48 @@ For each `gamign` ↔ `env` cluster:
 | funi | Aquamarine / turquoise | 7 |
 
 When adding markers via MCP: **new agent stamps** → red = 1; agent `INT:` / `LORE:` → white = 5. Do not change colors on markers you did not add.
+
+---
+
+## DRACONIANDISCOURSE S2 hybrid streams (Draconian Discourse + gaming)
+
+**Since S2**, most live sessions use a **hybrid timeline** — not the old split where gaming VOD and DD were separate exports.
+
+| Deliverable | Timeline | Description file | Notes |
+|-------------|----------|------------------|-------|
+| **Full stream VOD** | `{date} {name}.ts` — DD block **then** gaming block in one sequence | `{year}/editDesc/{date} {name}.ts.txt` | Same Premiere project as the live recording. Markers cover **both** DD and gaming. |
+| **DD-only upload** | Trimmed DD segment (same channel, same thumb) | `DRACONIANDISCOURSE/S2/E{n}/DD_S2E{n}.txt` stylized manifest | Derivative cut — **not** a second full timeline. Chapters come from the DD portion only. |
+| **DD-only stream** (rare) | DD only — no gaming tail | DD manifest only | When the whole session is discourse; no post-DD `gamign` block to stamp. |
+
+YouTube full VOD keeps **`starting soon`** from 0:00. RSS / DD derivative often begins at **`PRERAMBLE`** (first real content), not cold open.
+
+### Stamping style (from DD S2 workflow)
+
+Reference: `0STREAMUNIVERSE/DRACONIANDISCOURSE/.specstory/history` (S2E2–S2E5 marker passes).
+
+**DD portion (before first post-discourse `gamign` / `CONCLUSION` boundary):**
+
+1. **Topic tokens** — short `ALL_CAPS` labels, often `WORD_WORD` (e.g. `TRIP_RETROSPECTIVE`, `INTRO_HAWLEY`, `STEAM_HOOK`, `FUND_HYPOCRISY`). Match the tone of `DD_S2E{n}.txt` manifest chapters.
+2. **Tie to OBS times** — chapter rows in `editDesc` / DD manifest must use **marker timestamps**, not transcript guesses alone.
+3. **Preserve internal stamps** — `INT:` / converted `SEG`, all `FUNI`, pipeline `gamign`/`env`: keep names and colors; strip from deliverable chapter lists only.
+4. **Broad topics** — compress rapid-fire stamps within ~1 min into one label; don’t micro-chapter every tangent.
+5. **Bookends** — `starting soon` / `PRERAMBLE` / `goodbye` (or `GOODBYE` in DD manifest); last `env` → `postramble` in `editDesc` when applicable.
+
+**Gaming portion (after DD ends — first sustained `gamign` / `silksong prep` / equivalent):**
+
+1. **Rename & condense** purple (`gamign`/`env`), orange (`SEG`), and tan/yellow (`BEGIN`/`END`/`AFK`/`UNAFK`) stamps that remain after the DD cutoff — merge research toggles (`gamign`↔`env` ping-pong) into one deliverable label where possible.
+2. **Labels** — lowercase descriptive gameplay names in `{name}.ts.txt` (`silksong prep`, `lost lace`); agent **red** only for **new** location/boss/activity stamps the user hasn’t set.
+3. **Transcript + frames** — use `*_prtranscript.json` seeds plus ffmpeg probes (`_floom_probe/`, etc.) or `premiere_capture_frame_base64` for boss titles, map names, opsec clusters.
+4. **Do not** recolor or rename the user’s existing markers; agent adds red / white only.
+
+**Premiere hygiene (DD passes):**
+
+- **Never** `premiere_delete_all_markers` on hybrid sequences.
+- To replace DD chapter rows: ExtendScript delete markers **before** cutoff time only, then add chapter markers; leave post-`SEG` / post-`CONCLUSION` gaming markers intact.
+- Register undo (`premiere_undo`) before bulk rename/delete experiments.
+
+**Agent scope on hybrid `*.ts` sequences:**
+
+- Full `{name}.ts.txt` — semi-manual; agent does **not** auto-fill. Stamp Premiere + write `prtranscript/` sidecars.
+- `DD_S2E{n}.txt` — separate stylized manifest for the DD-only derivative; topic tokens only for the DD block.
+- Gaming agent pass runs on the **same** active sequence, scoped to timestamps **after** the DD boundary (see `editDesc` `CONCLUSION` / first post-DD `gamign`).