diff --git a/.claude/rules/ticket-cli.md b/.claude/rules/ticket-cli.md
index a54b7f856..997ef48d3 100644
--- a/.claude/rules/ticket-cli.md
+++ b/.claude/rules/ticket-cli.md
@@ -37,7 +37,7 @@ pql ticket board [--team T]                 # kanban view
 pql ticket status T-440 in_progress         # backlog|ready|in_progress|review|done|cancelled
 pql ticket assign T-440 dudley
 pql ticket team   T-440 server
-pql ticket setparent T-9 T-2                 # (clear with no parent arg)
+pql ticket setparent T-9 T-2                 # (clear with: setparent T-9 none)
 pql ticket append T-440 "extra context"      # also --file PATH / --stdin
 
 # Dependencies
diff --git a/.claude/skills/atlas/SKILL.md b/.claude/skills/atlas/SKILL.md
index e1c84293a..97e8dcfd8 100644
--- a/.claude/skills/atlas/SKILL.md
+++ b/.claude/skills/atlas/SKILL.md
@@ -4,7 +4,7 @@ description: >
   Query and manage celestial bodies and stations in systems.db. Use when the user
   says "atlas", "show system", "list bodies", "list stations", "author system",
   "show body", or invokes /atlas. Wraps the Rust atlas CLI binary. Also provides
-  the full per-system authoring pipeline via "author-system" subcommand.
+  the full per-system authoring pipeline via the /atlas author-system workflow.
 user-invocable: true
 allowed-tools: Bash, Read, Grep, Glob, Write, Edit, Agent
 ---
@@ -115,15 +115,13 @@ tooling/atlas stats
 ```
 Returns: system count, body count by type, inhabited count, station count by type.
 
-### Bulk populate from system data (classifier pass)
+### Corridor status (remaining unfinished systems)
 ```bash
-tooling/atlas populate --dry-run   # preview what would be created
-tooling/atlas populate              # create body/station records
+tooling/atlas corridor-status
 ```
-Reads `habitable_planet_count`, `inhabited_planet_count`, `gas_giant`,
-`asteroid_belt`, and `horizon_station` from existing system data.
-Creates: planets (inhabited first), gas giants, asteroid belts, oort cloud per
-system, horizon station per oort cloud. Skips systems that already have bodies.
+Shows remaining unfinished systems grouped by geographic sector and hop
+distance. Supersedes the old bulk-populate classifier pass (removed) — use
+the per-system `author` → review → `commit-system` workflow below instead.
 
 ### Update a system-level field
 ```bash
@@ -149,20 +147,33 @@ the commit message (defaults to `east_reach`).
 
 ## Body ID Naming Convention
 
+Real forms, verified against the ~3,300 IDs in `docs/atlas/proposals/`:
+
 ```
-GJ-{n}          — star (single)
-GJ-{n}A         — primary star (binary)
-GJ-{n}B         — secondary star (binary)
-GJ-{n}b/c/d...  — planets, innermost first
-GJ-{n}Ab/c/d... — planets orbiting primary only
-GJ-{n}Bb/c/d... — planets orbiting secondary only
-GJ-{n}d-1       — first moon of third planet
-GJ-{n}d-S1      — first station orbiting third planet
-GJ-{n}-oort     — oort cloud region
-GJ-{n}-belt     — asteroid belt
-GJ-{n}-oort-S1  — horizon station in oort cloud
+GJ {n}          — system_id (space, e.g. "GJ 105A"; "A"/"B" suffix for binary)
+GJ{n}A          — primary star's body prefix (no dash after GJ)
+GJ{n}Ab/c/d...  — planets orbiting primary, innermost first (no dash)
+GJ{n}Bb/c/d...  — planets orbiting secondary only
+GJ{n}Ae-m1      — first moon of body "e" (both forms seen — see below)
+GJ{n}Ae-1       — first moon of body "e" (bare-number form, also seen)
+GJ{n}Ac-S1      — first station orbiting body "c"
+GJ{n}-belt      — asteroid belt
+GJ{n}-oort      — oort cloud region
+GJ{n}-oort-S1   — horizon station in oort cloud
 ```
 
+**Moon suffixes are not standardized** — both `-m{n}` (e.g. `GJ105Ae-m1`) and
+bare `-{n}` (e.g. `GJ0d-1`) are in live use across proposals in roughly equal
+numbers (~208 vs ~228 of 436 moon entries as of this check). Match whichever
+form the system's existing proposal already uses; don't silently convert one
+to the other.
+
+**Wiki slugs use the dashed form** (`GJ-105A`, spaces→dashes) for the system
+directory under `wiki/star-systems/`, but body/station IDs inside that
+directory keep the undashed body-ID form (e.g.
+`wiki/star-systems/GJ-105A/bodies/GJ105Ac`) — the dash only applies to the
+system-level slug, never to body IDs.
+
 ## Entity Hierarchy
 
 ```
@@ -200,19 +211,32 @@ See `references/author-system.md` for the full workflow.
 - `references/review-ozzie.md` — Ozzie's review checklist (includes astrophysical sanity)
 - `references/integrity-check.md` — automated verification rules
 
-## Scope — A True Atlas
+## Asset Pipeline Relationship
 
-This skill will grow to cover the full geographic hierarchy of the Reach:
+See `.claude/rules/asset-pipeline.md` (the "Golden Rule": edit sources, not
+`systems.db`) and D-223. The bodies/stations catalog this skill authors is
+**atlas-CLI-owned** and survives `make regen-db`: `import_economics` only
+*enriches* existing body rows (radius, axial tilt, biosphere class — see
+`tooling/economy-db/economy_import/bodies.py`), it never deletes or
+regenerates the catalog. What `import_economics` *does* clear on every regen
+is the separate `atlas_*` geometry index tables (`atlas_cities`,
+`atlas_roads`, etc., D-223) — those hold cascade-computed geometry, not the
+hand-authored body/station identity data. So: **atlas write commands
+(`add-body`, `add-station`, `author`/`commit-system`) are the sanctioned path
+for the bodies/stations catalog**; they don't need a regen-db round-trip and
+aren't at risk from one.
 
-- **Galactic:** 301 systems, gate topology, sector/corridor data
-- **System:** orbital bodies, stations, oort cloud
-- **Planetary surface:** continents, oceans, mountain ranges, rivers, biome regions
-- **Settlement:** cities, towns, villages, outposts, rail lines, road hierarchy
-- **District:** neighborhoods, zones, named areas/provinces
-- **Local:** named locations, landmarks, facilities
+## Scope — Atlas Data Ownership
 
-As the development cascade progresses through Phase 1 (wiki content) → Phase 3
-(planetary maps / Atlas of the Reach), the atlas CLI and its schema will extend
-to cover each level. The goal is a single queryable geographic database from
-galaxy scale to street level — the in-game implant's atlas app reads from this
-same data.
+The atlas CLI owns the **galactic** and **system** levels — star systems,
+orbital bodies, stations — as hand-authored, reviewed data via this skill's
+`author-system` pipeline. That catalog is durable (see "Asset Pipeline
+Relationship" above).
+
+Everything below system scale — planetary surface, settlements, districts,
+local detail — is **not** grown by extending this CLI or its schema. D-223
+retired the Python atlas geometry generator; the deterministic server-side
+generation cascade (Phase 4, the active phase — see CLAUDE.md's Development
+Cascade) is the sole producer of that geometry going forward. The `atlas_*`
+index tables exist so the implant Atlas viewer can display cascade output,
+not as a target for this skill to populate by hand.
diff --git a/.claude/skills/atlas/references/integrity-check.md b/.claude/skills/atlas/references/integrity-check.md
index fc736fa76..4f90b0c37 100644
--- a/.claude/skills/atlas/references/integrity-check.md
+++ b/.claude/skills/atlas/references/integrity-check.md
@@ -2,48 +2,64 @@
 
 Run these checks on the proposal JSON before commit. All must pass.
 
+`tooling/atlas-verify` only ever opens the proposal JSON file — it never opens
+`systems.db`. Checks that require comparing against the live database (2, 4)
+are **not enforced by the script** and must be performed manually or by the
+authoring agent before commit.
+
 ## Checks
 
-1. **Inhabited bodies have names:**
+1. **Inhabited bodies have names** — *(script-checked)*
    Every body with `inhabited: true` must have `proper_name` set (not null).
 
-2. **Uninhabited bodies have no names:**
-   Every body with `inhabited: false` must have `proper_name: null`.
+2. **Uninhabited bodies have no names** — *(manual/agent-performed — the
+   script explicitly skips this check)*
+   Every body with `inhabited: false` should have `proper_name: null`.
    Exception: bodies that serve as named research outposts (e.g., "Corrosion
    Bench") where the name IS the body's identity.
 
-3. **All stations have names:**
+3. **All stations have names** — *(script-checked)*
    Every station must have `proper_name` set. Use the station_id as the
    name if the station has no cultural name (e.g., "GJ406g-S1" for an
    anonymous fuel depot).
 
-4. **No name collisions:**
+4. **No name collisions** — *(manual/agent-performed — requires querying
+   `systems.db`, which the script never opens)*
    ```sql
    SELECT proper_name FROM bodies WHERE proper_name IS NOT NULL
    UNION SELECT proper_name FROM stations WHERE proper_name IS NOT NULL
    ```
    No proposed name may duplicate an existing name in the database.
 
-5. **Body count minimum:**
+5. **Body count minimum** — *(script-checked)*
    - G/F stars: ≥ 8 planets (excluding moons, belts, oort)
    - K stars: ≥ 7 planets
-   - M dwarfs: ≥ 6 planets
-   - A/B stars: ≥ 6 planets
+   - M dwarfs and A/B stars: ≥ 6 planets (same threshold, script does not
+     distinguish these two cases)
 
-6. **Required structures:**
+6. **Required structures** — *(script-checked)*
    - Oort cloud body present
    - Horizon station present with `has_gate_infrastructure: true`
    - At least 1 asteroid belt (unless wiki explicitly contradicts)
 
-7. **Orbit consistency:**
+7. **Orbit consistency** — *(script-checked)*
    - `orbit_index` increases monotonically for top-level bodies
    - No duplicate `orbit_index` at the same parent level
-   - Moon `orbit_index` starts at 1 within parent
+   - Same monotonic + no-duplicate check applied per-parent to moon
+     `orbit_index` values (the script does not separately verify that moon
+     numbering starts at 1)
 
-8. **Parent references valid:**
+8. **Parent references valid** — *(script-checked)*
    - Every moon's `parent_body_id` references a body in the same proposal
    - Every station's `orbits_body_id` references a body in the same proposal
 
+9. **star_type vs spectral_class consistency** — *(script-checked, not
+   previously documented here)*
+   The leading letter of `star_type` (e.g. "K") must match the leading class
+   letter of `spectral_class` after stripping dwarf/subdwarf prefixes (e.g.
+   "K3V" → "K"). Skipped when `star_type` is `"binary"` or `"unusual"`, or
+   when either field is missing.
+
 ## Running the Check
 
 ```bash
@@ -57,3 +73,12 @@ make atlas-verify
 
 The script reports all failures per file, not just the first one.
 Exit code 1 if any file has errors.
+
+**Exclusions:** `GJ0.json` (Sol) and `GJ1221.json` are permanently excluded
+from the script's checks. Sol is hand-authored with different rules (named
+uninhabited bodies, real Earth/Mars/Luna geography) and is exempt from the
+normal generator/naming pipeline (D-223). GJ1221 is a dead-star remnant system
+("stellar consumption remnant" per its proposal notes) with zero planets —
+just a belt, an oort cloud, and a research horizon station — which would
+otherwise fail the body-count-minimum check (5) that assumes every system has
+planets.
diff --git a/.claude/skills/atlas/references/naming-brief.md b/.claude/skills/atlas/references/naming-brief.md
index 069cc91df..013f07c75 100644
--- a/.claude/skills/atlas/references/naming-brief.md
+++ b/.claude/skills/atlas/references/naming-brief.md
@@ -23,7 +23,8 @@ Collision avoidance — these names are already in use:
  UNION SELECT proper_name FROM stations WHERE proper_name IS NOT NULL}
 
 Name ALL inhabited bodies AND ALL stations in one pass.
-Uninhabited bodies do NOT get names (proper_name stays null).
+Uninhabited bodies do NOT get names (proper_name stays null) — unless the
+name IS the body's identity (a named lore outpost, e.g. "Corrosion Bench").
 
 ONE name + one-line rationale per location. Message only, no file.
 ```
@@ -38,4 +39,6 @@ ONE name + one-line rationale per location. Message only, no file.
   "The Exchange"). They should feel like real place names that happened to
   be assigned to transit/exchange facilities.
 - Every inhabited body and every station MUST get a proper_name.
-- Uninhabited bodies MUST have proper_name = null (body_id is the UI label).
+- Uninhabited bodies default to proper_name = null (body_id is the UI label) —
+  unless the name IS the body's identity (named lore outposts, e.g. "Corrosion
+  Bench"), matching the exception in `references/integrity-check.md` check 2.
diff --git a/.claude/skills/audio-gen/SKILL.md b/.claude/skills/audio-gen/SKILL.md
index 4c820e289..8d80072c7 100644
--- a/.claude/skills/audio-gen/SKILL.md
+++ b/.claude/skills/audio-gen/SKILL.md
@@ -19,6 +19,12 @@ Asset descriptions, filenames, bus routing, and design intent are documented in
 `docs/assets/audio/`. This skill provides the prompt system, generation
 workflow, and quality validation.
 
+`audio-health`, `audio-generate`, and `audio-batch` re-exec into the project
+`.venv` on startup (`audio-post` doesn't need to — it only shells out to
+ffmpeg). On a fresh clone with no `.venv` yet, they fail fast with `error:
+.venv not found — run make setup-venv first.` — run that once before using
+this skill.
+
 ## API Access
 
 **Never call the API directly.** Use the wrapper scripts:
@@ -39,7 +45,7 @@ tooling/db/audio-generate "prompt text" \
   --output-ogg client/assets/audio/final.ogg
 
 # Batch-generate from a manifest (preferred for multiple assets)
-tooling/db/audio-batch docs/assets/audio/batch-s10-327.json
+tooling/db/audio-batch docs/assets/audio/<manifest>.json
 ```
 
 ### Parameters
@@ -87,11 +93,12 @@ approvals to 2: one Write (manifest) + one Bash (batch run).
 
 ### 1. Create the manifest
 
-Write a JSON manifest to `docs/assets/audio/batch-{sprint}-{ticket}.json`:
+Write a JSON manifest to `docs/assets/audio/batch-T-<NNN>.json` (current
+ticket-ID scheme, e.g. `batch-T-327.json` — not the retired sprint scheme):
 
 ```json
 {
-  "description": "Sprint 10 ambient + world SFX batch",
+  "description": "T-327 ambient + world SFX batch",
   "output_dir": "client/assets/audio",
   "gen_dir": "client/assets/audio/gen",
   "defaults": {
@@ -134,20 +141,27 @@ Write a JSON manifest to `docs/assets/audio/batch-{sprint}-{ticket}.json`:
 Asset `id` values must match IDs in `docs/assets/audio/{category}.md` (e.g.,
 AMB-001, SFX-002, UI-005). This couples the manifest to the asset inventory.
 
+**`lufs`/`quality` defaults apply to `synth` assets only.** `audio_batch.py`'s
+SAO path (`run_sao_generate`) only forwards `steps`/`cfg`/`timeout` to the
+connector — `--lufs`/`--quality` aren't even exposed as CLI flags on
+`audio_connector.py generate`, so a manifest's `defaults.lufs`/`defaults.quality`
+are silently ignored for `method: "sao"` assets. SAO post-processing is fixed
+at -16 LUFS / quality 6 regardless of what the manifest says.
+
 ### 2. Run the batch
 
 ```bash
 # Full run
-tooling/db/audio-batch docs/assets/audio/batch-s10-327.json
+tooling/db/audio-batch docs/assets/audio/<manifest>.json
 
 # Dry run — preview what would be generated
-tooling/db/audio-batch docs/assets/audio/batch-s10-327.json --dry-run
+tooling/db/audio-batch docs/assets/audio/<manifest>.json --dry-run
 
 # Generate only specific assets
-tooling/db/audio-batch docs/assets/audio/batch-s10-327.json --only AMB-001,AMB-002
+tooling/db/audio-batch docs/assets/audio/<manifest>.json --only AMB-001,AMB-002
 
 # Skip assets that already have OGG files
-tooling/db/audio-batch docs/assets/audio/batch-s10-327.json --skip-existing
+tooling/db/audio-batch docs/assets/audio/<manifest>.json --skip-existing
 ```
 
 ### 3. Update asset docs with prompts
@@ -250,6 +264,9 @@ After generating, verify:
 ## File Placement
 
 Generated assets go to `client/assets/audio/` with exact filenames from the
-asset docs. Intermediates go to `client/assets/audio/gen/` (gitignored).
+asset docs. Intermediates go to `client/assets/audio/gen/` — untracked (the
+directory doesn't exist until first use and isn't in `.gitignore`); delete
+intermediates yourself after selecting the final candidate rather than
+relying on git to ignore them.
 
 AudioManager discovers assets by directory scan — filenames must match exactly.
diff --git a/.claude/skills/audio-gen/references/sonic-palette.md b/.claude/skills/audio-gen/references/sonic-palette.md
index d9d2751d8..829aee682 100644
--- a/.claude/skills/audio-gen/references/sonic-palette.md
+++ b/.claude/skills/audio-gen/references/sonic-palette.md
@@ -32,7 +32,10 @@ For environmental sounds, human cognition, physical world.
 **Character:** Human warmth. Natural attack and decay. Reverberant — exists
 in physical space. The sound of a real place with real materials.
 
-**Frequency range:** 60Hz-3kHz primary. Soft or natural attack, room tail.
+**Frequency range:** 300Hz-1.5kHz primary. Soft or natural attack, room tail.
+(The wider 60Hz-3kHz span belongs to the Station Baseline layered bed below —
+low-frequency foundation through high-frequency detail — not to a single
+Organic source sound.)
 
 **Used for:** fog_recognition, all amb_* ambient loops, all sfx_footstep_*
 
diff --git a/.claude/skills/bug-report/SKILL.md b/.claude/skills/bug-report/SKILL.md
index c176bc354..f14d0749b 100644
--- a/.claude/skills/bug-report/SKILL.md
+++ b/.claude/skills/bug-report/SKILL.md
@@ -6,7 +6,7 @@ description: >
   or invokes /bug-report. Scans the user:// bug-reports directory, summarizes
   each report, and offers investigation, ticket creation, or dismissal.
 user-invocable: true
-allowed-tools: Bash, Read, Grep, Glob, Write
+allowed-tools: Bash, Read, Grep, Glob
 ---
 
 # Bug Report Skill
@@ -26,12 +26,13 @@ contains these files:
 
 | File | Purpose |
 |------|---------|
-| `description.txt` | Tester notes + metadata (tick, room, stance, facing, position) |
+| `description.txt` | Tester notes + metadata (tick, room, stance, facing, position, timestamp, ring-buffer depth) |
 | `render.txt` | Simplified text render of the game snapshot |
+| `screenshot.png` | Screenshot of the game view at the moment the report was filed |
 | `snapshot.json` | Full JSON snapshot (entities, dialogue state, etc.) |
 | `inputs.jsonl` | Last 60 ticks of player input (replay format) |
 | `snapshots.jsonl` | Last 60 ticks of observer snapshots |
-| `seed.txt` | RNG seed for deterministic replay |
+| `seed.txt` | RNG seed for deterministic replay, or the literal string `unavailable` (see below) |
 
 ## Invocation
 
@@ -42,10 +43,13 @@ contains these files:
 
 ### 1. Scan for unprocessed reports
 
-List all report directories in the bug reports directory:
+List all report directories in the bug reports directory. Match the
+`gauntlet-*` naming pattern specifically, not a bare `*/` glob — this
+excludes the `archive/` subdirectory (see "Create ticket" below) from being
+mistaken for a report:
 
 ```bash
-ls -1d "/var/home/jeroenschweitzer/.local/share/godot/app_userdata/The Settled Reach/bug-reports/"*/
+ls -1d "/var/home/jeroenschweitzer/.local/share/godot/app_userdata/The Settled Reach/bug-reports/"gauntlet-*/
 ```
 
 If no directories are found, report "No bug reports found." and stop.
@@ -62,10 +66,17 @@ For each report directory, read the following files using the Read tool:
    - Tick number
    - Room name
    - Player stance, facing, position
+   - `Timestamp` (wall-clock time the report was filed)
+   - `RingBufferTicks` (depth of the `inputs.jsonl`/`snapshots.jsonl` history —
+     60 today, but read it rather than assume)
 2. **`render.txt`** — extract:
    - A brief description of what the text render shows (room layout, visible
      entities, player position marker)
-3. **`snapshot.json`** — extract:
+3. **`screenshot.png`** — view with the Read tool (it renders the image) and
+   note what it shows. Compare against `render.txt`/`snapshot.json` — a
+   mismatch between what the screenshot shows and what the snapshot data says
+   is itself a bug signal (see Tips: rendering vs simulation desync).
+4. **`snapshot.json`** — extract:
    - Total entity count
    - Whether dialogue is active (look for `dialogue` or `conversation` keys
      with non-null/non-empty values)
@@ -73,7 +84,11 @@ For each report directory, read the following files using the Read tool:
      non-empty values)
    - NPC names and positions if present
    - Any error or anomaly fields
-4. **`seed.txt`** — note the seed value for reference
+5. **`seed.txt`** — note the seed value for reference. If the file contains
+   the literal string `unavailable`, deterministic replay is not currently
+   possible (the server doesn't yet emit `rng_seed` on `ObserverSnapshot`) —
+   report "seed unavailable — replay not possible", don't paste the sentinel
+   into a ticket as if it were a usable seed.
 
 Do NOT read `inputs.jsonl` or `snapshots.jsonl` during the summary phase.
 These are large files reserved for the investigation step.
@@ -121,9 +136,13 @@ Dig deeper into the report for root cause analysis:
    - State machine transitions that look wrong
    - Component values going out of expected range
    - Entities appearing or disappearing unexpectedly
-4. Cross-reference with `render.txt` to confirm visual manifestation
-5. Read `seed.txt` and note it — the seed plus `inputs.jsonl` should allow
-   deterministic replay of the scenario
+4. Cross-reference with `render.txt` and `screenshot.png` to confirm visual
+   manifestation — a mismatch between the screenshot and the snapshot/render
+   data is itself the rendering-desync signal described in Tips
+5. Read `seed.txt` and note it — if it holds a real seed, it plus
+   `inputs.jsonl` should allow deterministic replay of the scenario; if it
+   holds the `unavailable` sentinel, say so explicitly and skip the replay
+   step rather than treating the sentinel as a seed
 
 Present findings as a root cause analysis:
 
@@ -143,8 +162,8 @@ Present findings as a root cause analysis:
 - {system 2}: {how it's involved}
 
 ### Reproduction
-Seed: {seed}
-Replay inputs.jsonl from tick {start} to reproduce.
+Seed: {seed, or "unavailable — replay not possible" if seed.txt held the sentinel}
+Replay inputs.jsonl from tick {start} to reproduce (seed permitting).
 
 ### Suggested fix
 {If identifiable from the snapshot data}
@@ -166,20 +185,42 @@ nature of the bug:
   issues)
 
 Construct the ticket title and description from the report summary and any
-investigation findings. Use the pql CLI:
+investigation findings. Parent the ticket per the cascade model (CLAUDE.md
+Development Cascade / `.claude/rules/ticket-cli.md`):
+- **`--parent T-1037`** for maintenance/test-debt bugs (bugs in test
+  infrastructure, tooling, or process hygiene rather than game systems)
+- **the active phase epic** for phase-scoped bugs (bugs in game systems
+  currently under construction) — don't hardcode the id; verify first with
+  `pql ticket show T-745 --tree` and use whichever phase epic has
+  `"status": "in_progress"` (currently `T-750`, Phase 4, but the cascade
+  advances)
+
+Use the pql CLI:
 
 ```bash
-pql ticket new bug "{title}" --team {team} --description "{description}"
+pql ticket new bug "{title}" --team {team} --description "{description}" --parent {T-1037 or the active phase epic}
 ```
 
 The description should include:
 - Bug summary (from tester notes)
 - Tick, room, position
 - Key observations from snapshot analysis
-- Seed for reproduction
+- Seed for reproduction, or "seed unavailable — replay not possible" if
+  `seed.txt` held the sentinel
 - Report directory name for reference
 
-After creating the ticket, report the ticket ID to the user.
+After creating the ticket, **verify it** with `pql ticket show <id>` (per
+`.claude/rules/ticket-cli.md`'s verify-after-create step) to confirm the
+title and parenting landed clean, then report the ticket ID to the user.
+
+A report is "processed" once it has a resulting ticket (or was dismissed) —
+after reporting the ticket ID, offer to either delete the report directory
+(same confirmation rule as Dismiss below) or move it to
+`{bug-reports-dir}/archive/{report-dir}/`, for cases where the tester or a
+reviewer might want to re-open it later. **Always confirm with the user
+first**, same as Dismiss. This is what makes "Remaining unprocessed" in the
+final summary computable — without it, a ticketed report is indistinguishable
+from a fresh one on the next run.
 
 #### Dismiss
 
@@ -226,6 +267,9 @@ After all reports have been processed, present a summary:
   positions in the render should match positions in the snapshot. Mismatches
   are themselves a bug signal (rendering vs simulation desync).
 - The seed in `seed.txt` combined with `inputs.jsonl` enables deterministic
-  replay on the server. Note this in any ticket you create.
+  replay on the server — when a real seed is present. Today `seed.txt` holds
+  the literal string `unavailable` in every report (the server doesn't yet
+  emit `rng_seed` on `ObserverSnapshot`); note whichever is true in any
+  ticket you create, don't assume replay is possible.
 - If the bug-reports directory does not exist, the tester has not yet run any
   gauntlet sessions or has not filed any reports. This is not an error.
diff --git a/.claude/skills/glb-gen/SKILL.md b/.claude/skills/glb-gen/SKILL.md
index d971c6a45..cc41c47e2 100644
--- a/.claude/skills/glb-gen/SKILL.md
+++ b/.claude/skills/glb-gen/SKILL.md
@@ -18,7 +18,7 @@ Convert approved concept images to game-ready .glb models.
 | Service | Check |
 |---------|-------|
 | Trellis | `tooling/db/trellis_connector.py health` |
-| Blender | `flatpak run org.blender.Blender --version` |
+| Blender | `tooling/blender --version` |
 
 Trellis runs on tower-of-joy and may be switched off. Check before batching.
 
@@ -41,54 +41,35 @@ python3 tooling/db/trellis_connector.py generate input.png \
 
 ## Batch Usage
 
-For multiple models, use the batch script:
+`tooling/trellis-batch.sh` exists but is **hardcoded to one category**: it
+takes no arguments and always processes the 16 character body types from
+`.tmp/image-gen/characters/bodies` into `.tmp/glb-gen/characters/bodies`.
+Running it for any other asset category (furniture, props, etc.) does
+nothing useful — it will just re-run (or skip, if outputs already exist) the
+same 16 character bodies regardless of what you intended.
+
+For any other category, loop `trellis_connector.py` calls yourself with a
+cooldown between jobs:
 
 ```bash
-bash tooling/trellis-batch.sh
+for f in .tmp/image-gen/furniture/tables/*.png; do
+  name=$(basename "$f" .png)
+  python3 tooling/db/trellis_connector.py generate "$f" \
+    --output ".tmp/glb-gen/furniture/tables/${name}.glb" \
+    --simplify 0.95 --texture-size 1024
+  sleep 15
+done
 ```
 
-The batch script runs one model at a time with 15s cooldown between jobs,
-retries up to 3 times on failure, and skips already-generated GLBs.
-
 **Never run Trellis jobs in parallel** — it uses the full GPU and concurrent
 jobs will OOM and corrupt the CUDA state.
 
 ## Trellis API Reference
 
-The connector talks to a Gradio API. The parameter layout is fragile — document
-changes here when the container is updated.
-
-### /image_to_3d — 9 inputs
-
-| Pos | Name | Type | Value |
-|-----|------|------|-------|
-| 0 | image | Image | preprocessed image from /preprocess_image_1 |
-| 1 | multiimages | Gallery | `[]` for single-image mode |
-| 2 | is_multiimage | State | `False` for single-image (**not a session — it's a boolean flag**) |
-| 3 | seed | Slider | int, 0-2147483647 |
-| 4 | ss_guidance | Slider | float, default 7.5 |
-| 5 | ss_steps | Slider | int, default 12 |
-| 6 | slat_guidance | Slider | float, default 3.0 |
-| 7 | slat_steps | Slider | int, default 12 |
-| 8 | multiimage_algo | Radio | `"stochastic"` or `"multidiffusion"` |
-
-### /extract_glb — 3 inputs
-
-| Pos | Name | Type | Value |
-|-----|------|------|-------|
-| 0 | output_buf | State | `None` — server uses internal state from image_to_3d |
-| 1 | simplify | Slider | float, mesh simplification ratio (default 0.95) |
-| 2 | texture_size | Slider | int, texture resolution (default 1024) |
-
-### Common Failure Modes
-
-| Error | Cause | Fix |
-|-------|-------|-----|
-| "needed 9, got 8" on image_to_3d | Missing `is_multiimage` at position 2 | Pass `False` |
-| "needed 3, got 2" on extract_glb | Missing `output_buf` at position 0 | Pass `None` |
-| `'float' cannot be interpreted as int` | numpy rejects float for linspace steps | Patch `flow_euler.py` on server: `int(steps)` |
-| CUDA device mismatch / invalid argument | GPU state corrupted after crash | Restart the Trellis container |
-| Repeated 500 after restart | Previous crash left tensors on wrong device | Full container restart (not just API restart) |
+The connector talks to a fragile Gradio API (positional args, no field
+names). Full parameter tables and common failure modes are in
+`references/trellis-api.md` — read it when a Trellis call fails with a
+position/count mismatch or a CUDA/device error.
 
 ## Intermediate and Output Directories
 
@@ -112,7 +93,13 @@ Always mirror the category/subcategory path across stages so you can trace
 `image-gen/furniture/tables/foo_concept.png` → `glb-gen/furniture/tables/foo.glb`.
 
 Final game-ready assets are copied to `spikes/3dpipeline/models/` for spike
-testing, or to `client/assets/models/` when ready for production.
+testing. For production, character GLBs go to `client/assets/characters/`
+(already in active use — see its `bodies/`, `clothing/`, `hair/`,
+`skeleton/`, etc. subdirectories). `client/assets/models/` does not exist —
+there is no established production path yet for non-character categories
+(furniture, props); that's a Phase-5 concern (Player control & in-world
+rendering, per CLAUDE.md's Development Cascade). Until then, keep
+non-character output staged under `.tmp/glb-gen/postproc/`.
 
 ## Post-process in Blender
 
@@ -125,7 +112,7 @@ Normalize scale, center, generate recolor mask, adjust materials:
 Or directly:
 
 ```bash
-flatpak run org.blender.Blender --background \
+tooling/blender --background \
   --python .claude/skills/glb-gen/scripts/postprocess_glb.py \
   -- input.glb output.glb [--target-width N] [--color-threshold N]
 ```
@@ -139,7 +126,10 @@ For best Trellis results:
 - No text, labels, or watermarks
 - 3/4 front view at eye level gives Trellis the most information
 - **Do NOT force isometric angle** — Trellis reconstructs full 3D, the game camera handles the view
-- Use a style anchor image (`--input`) to maintain consistency across batches
+- To maintain consistency across a batch, generate the concept images with
+  `/image-gen` using its `--input` style anchor flag — `--input` is an
+  /image-gen flag, not a Trellis one; `trellis_connector.py` has no `--input`
+  argument and exits on unrecognized flags.
 
 These match `/image-gen` output with the Settled Reach style guide.
 
@@ -159,6 +149,16 @@ Set in `project.godot` under `[gltf]` so new imports pick it up automatically.
 
 ## Material Convention
 
-Post-processed .glb files use standard material slot names:
-`mat_wood_primary`, `mat_metal_primary`, `mat_fabric_primary`, etc.
-See `docs/design/character-visuals-spec.md` §5 for character materials.
+`scripts/postprocess_glb.py` renames **every** material slot on the imported
+mesh to a single slot, `mat_primary` by default (override with
+`--material-name NAME`) — it does not produce per-material names like
+`mat_wood_primary`/`mat_metal_primary`/`mat_fabric_primary`; those names are
+not used anywhere in this pipeline. The script also keeps the original
+Trellis-baked texture and sets roughness=1.0/specular=0.0 for toon
+compatibility (see `setup_materials()`).
+
+Note `docs/design/character-visuals-spec.md` §5 ("Color Mesh Regions") is a
+**different, unrelated system**: it defines 2D compositor color regions for
+the character-visuals rendering pipeline (`skin_primary`, `hair_primary`,
+`cloth_primary`/`cloth_secondary`/`cloth_accent`, `accessory_primary`, etc.),
+not GLB material slot names — do not conflate the two when naming things.
diff --git a/.claude/skills/glb-gen/references/trellis-api.md b/.claude/skills/glb-gen/references/trellis-api.md
new file mode 100644
index 000000000..85830de50
--- /dev/null
+++ b/.claude/skills/glb-gen/references/trellis-api.md
@@ -0,0 +1,40 @@
+# Trellis API Reference
+
+The connector (`tooling/db/trellis_connector.py`) talks to a Gradio API. The
+parameter layout is fragile — document changes here when the container is
+updated. This is the canonical copy; the connector's module docstring carries
+a duplicate for at-a-glance reference when reading the script directly — keep
+both in sync if you change either, but treat this file as the source of truth
+for the skill.
+
+## /image_to_3d — 9 inputs
+
+| Pos | Name | Type | Value |
+|-----|------|------|-------|
+| 0 | image | Image | preprocessed image from /preprocess_image_1 |
+| 1 | multiimages | Gallery | `[]` for single-image mode |
+| 2 | is_multiimage | State | `False` for single-image (**not a session — it's a boolean flag**) |
+| 3 | seed | Slider | int, 0-2147483647 |
+| 4 | ss_guidance | Slider | float, default 7.5 |
+| 5 | ss_steps | Slider | int, default 12 |
+| 6 | slat_guidance | Slider | float, default 3.0 |
+| 7 | slat_steps | Slider | int, default 12 |
+| 8 | multiimage_algo | Radio | `"stochastic"` or `"multidiffusion"` |
+
+## /extract_glb — 3 inputs
+
+| Pos | Name | Type | Value |
+|-----|------|------|-------|
+| 0 | output_buf | State | `None` — server uses internal state from image_to_3d |
+| 1 | simplify | Slider | float, mesh simplification ratio (default 0.95) |
+| 2 | texture_size | Slider | int, texture resolution (default 1024) |
+
+## Common Failure Modes
+
+| Error | Cause | Fix |
+|-------|-------|-----|
+| "needed 9, got 8" on image_to_3d | Missing `is_multiimage` at position 2 | Pass `False` |
+| "needed 3, got 2" on extract_glb | Missing `output_buf` at position 0 | Pass `None` |
+| `'float' cannot be interpreted as int` | numpy rejects float for linspace steps | Patch `flow_euler.py` on server: `int(steps)` |
+| CUDA device mismatch / invalid argument | GPU state corrupted after crash | Restart the Trellis container |
+| Repeated 500 after restart | Previous crash left tensors on wrong device | Full container restart (not just API restart) |
diff --git a/.claude/skills/glb-gen/scripts/postprocess b/.claude/skills/glb-gen/scripts/postprocess
index a6d48929a..235818d30 100755
--- a/.claude/skills/glb-gen/scripts/postprocess
+++ b/.claude/skills/glb-gen/scripts/postprocess
@@ -68,10 +68,28 @@ echo "Post-processing: $(basename "$INPUT_ABS")"
 echo "  Input:  $INPUT_REL"
 echo "  Output: ${OUTPUT_ABS#$(cd "$PROJECT_ROOT" && pwd)/}"
 
-flatpak run org.blender.Blender --background \
+LOG="${OUTPUT_ABS%.glb}_blender.log"
+
+# Blender does not reliably set a non-zero exit code on an uncaught Python
+# exception (--python-exit-code is not set), so failures must be detected
+# from the log content and from whether the output file actually appeared —
+# not from Blender's own exit status alone. Capture the full log first
+# (temporarily disabling -e so a non-zero Blender exit doesn't abort before
+# we can inspect it), then filter for the console summary.
+set +e
+"$PROJECT_ROOT/tooling/blender" --background \
     --python "$BLENDER_SCRIPT" \
-    -- "$INPUT_ABS" "$OUTPUT_ABS" "$@" 2>&1 \
-    | grep -E "^  |^=|WARNING|ERROR|Dominant|Mask"
+    -- "$INPUT_ABS" "$OUTPUT_ABS" "$@" > "$LOG" 2>&1
+BLENDER_EXIT=$?
+set -e
+
+grep -iE "^  |^=|warning|error|traceback|dominant|mask" "$LOG" || true
+
+if [ "$BLENDER_EXIT" -ne 0 ] || [ ! -f "$OUTPUT_ABS" ]; then
+    echo "ERROR: Blender post-process failed (exit $BLENDER_EXIT, output present: $([ -f "$OUTPUT_ABS" ] && echo yes || echo no))" >&2
+    echo "  Full log: ${LOG#$(cd "$PROJECT_ROOT" && pwd)/}" >&2
+    exit 1
+fi
 
 MASK="${OUTPUT_ABS%.glb}_mask.png"
 if [ -f "$MASK" ]; then
diff --git a/.claude/skills/glb-gen/scripts/postprocess_glb.py b/.claude/skills/glb-gen/scripts/postprocess_glb.py
index a238e930d..a0c6a5a7c 100644
--- a/.claude/skills/glb-gen/scripts/postprocess_glb.py
+++ b/.claude/skills/glb-gen/scripts/postprocess_glb.py
@@ -3,7 +3,7 @@
 GLB post-processor for The Settled Reach asset pipeline.
 
 Run via Blender headless:
-    flatpak run org.blender.Blender --background --python postprocess_glb.py -- input.glb output.glb [options]
+    tooling/blender --background --python postprocess_glb.py -- input.glb output.glb [options]
 
 Operations:
     1. Normalize scale to fit a target bounding box (default 1x1x1 world units)
@@ -42,7 +42,7 @@ def get_script_args():
 def parse_args(args):
     """Parse script arguments."""
     if len(args) < 2:
-        print("Usage: postprocess_glb.py -- input.glb output.glb [--target-width N] [--color #hex] [--material-name name]")
+        print("Usage: postprocess_glb.py -- input.glb output.glb [--target-width N] [--target-height N] [--color-threshold N] [--material-name name]")
         sys.exit(1)
 
     result = {
diff --git a/.claude/skills/image-gen/SKILL.md b/.claude/skills/image-gen/SKILL.md
index 2f615dd80..5e61e6217 100644
--- a/.claude/skills/image-gen/SKILL.md
+++ b/.claude/skills/image-gen/SKILL.md
@@ -10,21 +10,27 @@ description: >
 
 # Image Generation — The Settled Reach
 
-Generate concept images via the Gemini API. Standalone skill — does not chain
-into other pipelines. Output is a PNG file.
+Generate concept images via the Gemini API. Does not auto-invoke downstream
+skills — /glb-gen consumes its output from `.tmp/image-gen/` when you choose
+to feed a concept image into the 3D pipeline, but that's a manual next step,
+not something this skill triggers itself. Output is a PNG file.
 
 ## Prerequisites
 
 Requires `GEMINI_API_KEY` in environment (set in `.claude/settings.local.json`).
 
+Uses the canonical connector at `tooling/db/image_connector.py` (see
+`.claude/rules/project-structure.md` — `tooling/db/` is the documented home
+for asset/connector scripts; this skill has no local fork of it).
+
 ```bash
-python3 .claude/skills/image-gen/scripts/image_connector.py health
+python3 tooling/db/image_connector.py health
 ```
 
 ## Usage
 
 ```bash
-python3 .claude/skills/image-gen/scripts/image_connector.py generate \
+python3 tooling/db/image_connector.py generate \
   "prompt text" \
   --output path/to/output.png \
   --aspect 1:1
@@ -53,8 +59,8 @@ Always use `--output .tmp/image-gen/[category]/[subcategory]/[name]_concept.png`
 | Flag | Default | Description |
 |------|---------|-------------|
 | `--output` | auto-named in `~/Pictures/mcp-images/` | Output PNG path (use `.tmp/image-gen/...` instead) |
-| `--aspect` | `1:1` | Aspect ratio (1:1, 16:9, 3:2, 4:3, etc.) |
-| `--size` | standard | Resolution hint (1K, 2K, 4K) |
+| `--aspect` | `1:1` | Real API parameter (`generationConfig.imageConfig.aspectRatio`). Valid: `1:1`, `3:2`, `2:3`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9` |
+| `--size` | standard | Resolution **hint** only — appended to the prompt text, not a real API parameter; Gemini may not honor it |
 | `--input` | none | Input image for image-to-image variation |
 
 ## Art Style Guide — The Settled Reach
@@ -81,7 +87,8 @@ For concept images intended as Trellis input (3D model generation):
 ## Batch Usage
 
 Generate multiple variants by running the script multiple times with different
-prompts or seeds. Each call is independent — no state between calls.
+prompts. Each call is independent — no state between calls, and there is no
+`--seed` flag (the connector rejects unknown arguments).
 
 Show generated images to the user for review before proceeding to any
 downstream pipeline (e.g. /glb-gen).
diff --git a/.claude/skills/image-gen/scripts/image_connector.py b/.claude/skills/image-gen/scripts/image_connector.py
deleted file mode 100755
index 6c0884991..000000000
--- a/.claude/skills/image-gen/scripts/image_connector.py
+++ /dev/null
@@ -1,248 +0,0 @@
-#!/usr/bin/env python3
-"""
-Gemini image generator connector — direct API wrapper.
-
-Generates images via Google's Gemini 2.0 Flash image generation API.
-API key from GEMINI_API_KEY env var or config.json.
-
-Usage:
-    python3 image_connector.py health
-    python3 image_connector.py generate "prompt" [--output file.png] [--aspect 1:1] [--size 1K] [--input image.png]
-"""
-
-import base64
-import json
-import os
-import sys
-import urllib.error
-import urllib.request
-
-CONFIG_PATH = os.path.join(os.path.dirname(__file__), "config.json")
-DEFAULT_OUTPUT_DIR = os.path.expanduser("~/Pictures/mcp-images")
-
-
-def get_api_key():
-    """Get Gemini API key from env or config."""
-    key = os.environ.get("GEMINI_API_KEY")
-    if key:
-        return key
-    try:
-        with open(CONFIG_PATH) as f:
-            config = json.load(f)
-            return config.get("gemini_api_key", "")
-    except Exception:
-        pass
-    print(json.dumps({
-        "ok": False,
-        "error": "No GEMINI_API_KEY found in environment or config.json"
-    }, indent=2))
-    sys.exit(1)
-
-
-def health():
-    """Check if the Gemini API is reachable with the configured key."""
-    key = get_api_key()
-    url = f"https://generativelanguage.googleapis.com/v1beta/models?key={key}"
-    try:
-        req = urllib.request.Request(url, method="GET")
-        with urllib.request.urlopen(req, timeout=10) as resp:
-            data = json.loads(resp.read())
-            models = [m.get("name", "") for m in data.get("models", [])
-                      if "imagen" in m.get("name", "").lower()
-                      or "flash" in m.get("name", "").lower()]
-            print(json.dumps({
-                "ok": True,
-                "api": "gemini",
-                "image_capable_models": models[:5],
-            }, indent=2))
-    except Exception as e:
-        print(json.dumps({
-            "ok": False,
-            "error": str(e)
-        }, indent=2))
-        sys.exit(1)
-
-
-def generate(prompt, output=None, aspect_ratio="1:1", image_size=None,
-             input_image=None):
-    """
-    Generate an image from a text prompt using Gemini.
-
-    Args:
-        prompt: Text description of the image to generate
-        output: Output file path (default: auto-named in ~/Pictures/mcp-images/)
-        aspect_ratio: Aspect ratio (1:1, 16:9, 3:2, etc.)
-        image_size: Resolution hint (1K, 2K, 4K) - may not be honored
-        input_image: Optional input image path for image-to-image generation
-    """
-    key = get_api_key()
-
-    # Gemini image generation model
-    model = "gemini-2.5-flash-image"
-    url = f"https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={key}"
-
-    if output is None:
-        safe = "".join(c if c.isalnum() or c in "-_ " else "" for c in prompt[:40])
-        safe = safe.strip().replace(" ", "_").lower()
-        os.makedirs(DEFAULT_OUTPUT_DIR, exist_ok=True)
-        output = os.path.join(DEFAULT_OUTPUT_DIR, f"{safe}.png")
-
-    # Build the request
-    parts = []
-
-    # Add input image if provided (image-to-image)
-    if input_image:
-        if not os.path.isfile(input_image):
-            print(json.dumps({"ok": False, "error": f"Input image not found: {input_image}"}), indent=2)
-            sys.exit(1)
-        with open(input_image, "rb") as f:
-            image_data = base64.b64encode(f.read()).decode("utf-8")
-        # Detect mime type
-        ext = os.path.splitext(input_image)[1].lower()
-        mime = {"png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg",
-                ".webp": "image/webp"}.get(ext, "image/png")
-        parts.append({
-            "inlineData": {
-                "mimeType": mime,
-                "data": image_data
-            }
-        })
-
-    # Build enhanced prompt with aspect ratio and size hints
-    enhanced_prompt = prompt
-    if aspect_ratio and aspect_ratio != "1:1":
-        enhanced_prompt += f" Aspect ratio: {aspect_ratio}."
-    if image_size:
-        enhanced_prompt += f" Resolution: {image_size}."
-
-    parts.append({"text": enhanced_prompt})
-
-    payload = json.dumps({
-        "contents": [{"parts": parts}],
-        "generationConfig": {
-            "responseModalities": ["TEXT", "IMAGE"],
-        }
-    })
-
-    req = urllib.request.Request(
-        url,
-        data=payload.encode(),
-        headers={"Content-Type": "application/json"},
-        method="POST"
-    )
-
-    print(f"Generating image...", file=sys.stderr)
-    print(f"  Prompt: {prompt}", file=sys.stderr)
-    if input_image:
-        print(f"  Input image: {input_image}", file=sys.stderr)
-
-    try:
-        with urllib.request.urlopen(req, timeout=120) as resp:
-            result = json.loads(resp.read())
-    except urllib.error.HTTPError as e:
-        body = e.read().decode("utf-8", errors="replace")
-        print(json.dumps({
-            "ok": False,
-            "error": f"API error {e.code}: {e.reason}",
-            "details": body[:500]
-        }, indent=2))
-        sys.exit(1)
-    except Exception as e:
-        print(json.dumps({"ok": False, "error": str(e)}), indent=2)
-        sys.exit(1)
-
-    # Extract image data from response
-    candidates = result.get("candidates", [])
-    if not candidates:
-        print(json.dumps({
-            "ok": False,
-            "error": "No candidates in response",
-            "response": json.dumps(result)[:500]
-        }, indent=2))
-        sys.exit(1)
-
-    image_saved = False
-    text_response = ""
-
-    for candidate in candidates:
-        content = candidate.get("content", {})
-        for part in content.get("parts", []):
-            if "inlineData" in part:
-                # Image data
-                image_b64 = part["inlineData"]["data"]
-                image_bytes = base64.b64decode(image_b64)
-                os.makedirs(os.path.dirname(os.path.abspath(output)), exist_ok=True)
-                with open(output, "wb") as f:
-                    f.write(image_bytes)
-                image_saved = True
-            elif "text" in part:
-                text_response += part["text"]
-
-    if not image_saved:
-        print(json.dumps({
-            "ok": False,
-            "error": "No image data in response",
-            "text_response": text_response[:500],
-            "response": json.dumps(result)[:500]
-        }, indent=2))
-        sys.exit(1)
-
-    file_size = os.path.getsize(output)
-    print(json.dumps({
-        "ok": True,
-        "file": output,
-        "size_bytes": file_size,
-        "prompt": prompt,
-        "aspect_ratio": aspect_ratio,
-    }, indent=2))
-
-
-def main():
-    if len(sys.argv) < 2:
-        print("Usage:")
-        print("  image_connector.py health")
-        print("  image_connector.py generate 'prompt' [--output file.png] [--aspect 1:1] [--size 1K] [--input image.png]")
-        sys.exit(1)
-
-    cmd = sys.argv[1]
-
-    if cmd == "health":
-        health()
-    elif cmd == "generate":
-        if len(sys.argv) < 3:
-            print("Error: prompt required", file=sys.stderr)
-            sys.exit(1)
-
-        prompt = sys.argv[2]
-        output = None
-        aspect_ratio = "1:1"
-        image_size = None
-        input_image = None
-
-        i = 3
-        while i < len(sys.argv):
-            if sys.argv[i] == "--output" and i + 1 < len(sys.argv):
-                output = sys.argv[i + 1]
-                i += 2
-            elif sys.argv[i] == "--aspect" and i + 1 < len(sys.argv):
-                aspect_ratio = sys.argv[i + 1]
-                i += 2
-            elif sys.argv[i] == "--size" and i + 1 < len(sys.argv):
-                image_size = sys.argv[i + 1]
-                i += 2
-            elif sys.argv[i] == "--input" and i + 1 < len(sys.argv):
-                input_image = sys.argv[i + 1]
-                i += 2
-            else:
-                print(f"Unknown argument: {sys.argv[i]}", file=sys.stderr)
-                sys.exit(1)
-
-        generate(prompt, output=output, aspect_ratio=aspect_ratio,
-                 image_size=image_size, input_image=input_image)
-    else:
-        print(f"Unknown command: {cmd}", file=sys.stderr)
-        sys.exit(1)
-
-
-if __name__ == "__main__":
-    main()
diff --git a/.claude/skills/sprite-gen/SKILL.md b/.claude/skills/sprite-gen/SKILL.md
index 63e5406f1..81653dae1 100644
--- a/.claude/skills/sprite-gen/SKILL.md
+++ b/.claude/skills/sprite-gen/SKILL.md
@@ -23,6 +23,12 @@ artwork via the Godot render pipeline at `renderer/`.
 models to 4-direction sprites for a 2D view — that role is **retired** (D-244). It now
 serves flat 2D artwork only.
 
+## Authoring a New Scene
+
+To author a new scene, read `renderer/README.md` — "Adding a New Model" +
+Material Guidelines (`MeshInstance3D` root centered at origin, flat
+`StandardMaterial3D`, no emission/specularity).
+
 ## Render Pipeline
 
 ```bash
@@ -32,16 +38,22 @@ serves flat 2D artwork only.
 Scenes live at `renderer/models/<name>.tscn`. The script still emits the legacy spike
 layout — 12 PNGs in `renderer/output/` (4 cardinal directions × 3 resolutions:
 1024/256/64). For **flat artwork** the relevant output is the single front (south)
-view; the multi-direction output is legacy object-sprite behaviour. A flat-artwork
-(single-view) mode for the render script is the natural follow-up.
+view; the multi-direction output is legacy object-sprite behaviour.
 
 ### After rendering
 
 1. Inspect the output PNG — does it read as the intended artwork at in-world scale?
-2. Promote the final PNG to `client/assets/` as the surface texture/decal.
+2. Promote to `client/assets/textures/` or `client/assets/artwork/` (create on
+   first use) — **never `client/assets/sprites/`**, that path is the retired
+   object-sprite catalog (D-244). Register the promoted asset in a manifest
+   per D-241's promotion boundary (precedent: `client/assets/characters/manifest.json`).
 
 ### Troubleshooting
 
-- **No output files:** Godot needs a display. If running headless, prefix with `xvfb-run`.
+- **No output files / "No DISPLAY set":** Godot needs a display; the script
+  now detects a missing `DISPLAY` itself and prints the remedy. `xvfb-run` is
+  **not installed by default** on the dev host — install it first (e.g.
+  `xorg-x11-server-Xvfb` on Fedora/RHEL, `xvfb` on Debian/Ubuntu), then
+  re-run under `xvfb-run -a`.
 - **Scene not found:** check the `.tscn` exists in `renderer/models/`.
 - **Godot not found:** pass the Godot path as the second arg: `render.sh <name> /path/to/godot`.
diff --git a/.claude/skills/sprite-gen/scripts/render.sh b/.claude/skills/sprite-gen/scripts/render.sh
index f015c90ec..6f8c74421 100755
--- a/.claude/skills/sprite-gen/scripts/render.sh
+++ b/.claude/skills/sprite-gen/scripts/render.sh
@@ -24,24 +24,48 @@ echo "Rendering: $MODEL_NAME"
 echo "Godot: $GODOT"
 echo "Project: $PROJECT"
 
+# The render step (unlike --import) needs a display. Fail fast with a clear
+# remedy instead of letting Godot die on a cryptic low-level GL/X error.
+if [ -z "${DISPLAY:-}" ]; then
+    if command -v xvfb-run >/dev/null 2>&1; then
+        echo "No DISPLAY set — re-run under xvfb-run, e.g.:"
+        echo "  xvfb-run -a $0 $MODEL_NAME${2:+ $2}"
+        exit 1
+    else
+        echo "ERROR: No DISPLAY set and xvfb-run is not installed."
+        echo "  Install it (e.g. 'xorg-x11-server-Xvfb' on Fedora/RHEL, 'xvfb' on"
+        echo "  Debian/Ubuntu), then re-run under xvfb-run -a."
+        exit 1
+    fi
+fi
+
 # Ensure resources are imported (headless, no display needed)
 echo "Importing resources..."
 "$GODOT" --path "$PROJECT" --headless --import 2>&1 || echo "Warning: import step had errors (may be non-fatal)"
 
+# Clear stale output from any previous run of this model first, so a failed
+# render can't be masked by leftover files from an earlier successful run.
+rm -f "$OUTPUT/${MODEL_NAME}"_*.png
+
 # Run Godot with the render scene, passing model name as user arg
 "$GODOT" --path "$PROJECT" "$SCENE" -- "$MODEL_NAME" 2>&1
 
 # Check output
-EXPECTED_FILES=0
+FOUND_FILES=0
 for dir in north east south west; do
     for res in 1024 256 64; do
         f="$OUTPUT/${MODEL_NAME}_${dir}_${res}.png"
         if [ -f "$f" ]; then
-            EXPECTED_FILES=$((EXPECTED_FILES + 1))
+            FOUND_FILES=$((FOUND_FILES + 1))
         fi
     done
 done
 
 echo ""
-echo "Generated $EXPECTED_FILES/12 files in $OUTPUT/"
+echo "Generated $FOUND_FILES/12 files in $OUTPUT/"
 ls -la "$OUTPUT/${MODEL_NAME}"_*.png 2>/dev/null || echo "No output files found."
+
+if [ "$FOUND_FILES" -eq 0 ]; then
+    echo "ERROR: Godot produced no output files — render failed." >&2
+    exit 1
+fi
diff --git a/.claude/skills/ticket/SKILL.md b/.claude/skills/ticket/SKILL.md
index ff2998fb6..4b8a4e44c 100644
--- a/.claude/skills/ticket/SKILL.md
+++ b/.claude/skills/ticket/SKILL.md
@@ -1,9 +1,12 @@
 ---
 name: ticket
 description: >
-  Manage project tickets in the pql planning store. Use when the user says
-  "ticket", "create a ticket", "show tickets", "board", "phase", or invokes
-  /ticket. Wraps the pql CLI for structured project management.
+  Manage project tickets in the pql planning store — day-to-day ticket CLI
+  (list, create, update, board, phase epics). Use when the user says "ticket",
+  "create a ticket", "show tickets", "board", "phase epic", "what's in the
+  active phase", or invokes /ticket. Wraps the pql CLI for structured project
+  management. For changelog/planning-store internals (branch moves, recovery,
+  write-through mechanics), use the pql skill instead.
 user-invocable: true
 allowed-tools: Bash, Read, Grep, Glob
 ---
@@ -16,58 +19,22 @@ Manage tickets via **pql** (`.pql/pql.db`, rebuilt from the git-tracked
 
 ## Commands
 
-### List / browse
-```bash
-pql ticket list [--status S] [--team T] [--label L] [--assigned A] [--decision D-NNN] \
-                [--unblocked] [--leaf] [--under T-N]
-pql ticket board [--team T]              # kanban view, grouped by status
-pql ticket show T-440 [--with-context] [--with-children] [--with-blockers] [--tree]
-pql ticket show T-1,T-2,T-3              # batch
-```
-(pql ticket `list` has no `--type`/`--priority` filter — list and read the JSON, or use
-`--under <epic>` for an epic's subtree.)
+List/show/create/update/dependencies/labels/refinement syntax is **not
+repeated here** — see the full command reference linked above
+(`.claude/rules/ticket-cli.md`), the one canonical copy.
 
-### Create
-```bash
-pql ticket new <type> <title> [--parent T-N] [--priority P] [--decision D-NNN] [--team T] [--description TEXT] [--assign A]
-```
-Types: `initiative`, `epic`, `story`, `task`, `bug`. Priorities: `critical`, `high`, `medium`, `low`.
-
-### Update
-```bash
-pql ticket status T-440 in_progress      # backlog|ready|in_progress|review|done|cancelled (batch with commas)
-pql ticket assign T-440 dudley
-pql ticket team   T-440 server           # comma teams allowed: server,client
-pql ticket setparent T-9 T-2
-pql ticket append T-440 "extra context"  # --file PATH / --stdin
-pql ticket label  T-440 add|rm <label>
-```
-
-### Dependencies
-```bash
-pql ticket show T-440 --with-blockers     # what blocks this
-pql ticket block   T-7 --by T-440         # T-440 blocks T-7
-pql ticket unblock T-7 --from T-440
-```
+**Negative knowledge (not in the rule file):** `pql ticket list` has **no**
+`--type`/`--priority` filter — list and read the JSON, or use `--under
+<epic>` for an epic's subtree.
 
 ### Phases (the cascade hierarchy, not labels or a milestone entity)
 ```bash
-pql ticket list --under T-750 --unblocked       # ready work in the active phase (Phase 4)
+pql ticket list --under T-750 --unblocked       # ready work in the active phase (currently T-750, Phase 4)
 pql ticket setparent T-990 T-750                # put a ticket in the phase (re-parent)
 pql ticket show T-745 --tree                     # the cascade: initiative -> 6 phase epics
 ```
 The active phase is the `in_progress` phase epic under T-745 (sequential per D-166).
 
-### Refinement & planning
-```bash
-pql ticket refine list                    # tickets with no description
-pql ticket refine next [--skip N]
-pql ticket refine write T-5 '{"description":"..."}'
-pql plan whatsnext                        # next ticket to work on
-pql plan review                           # next ticket awaiting review
-pql plan status                           # decision + ticket dashboard / counts
-```
-
 ## Workflow
 
 1. Decisions from `governance/` records become **initiatives**
@@ -77,3 +44,6 @@ pql plan status                           # decision + ticket dashboard / counts
 5. **Phase gating** is the cascade hierarchy (initiative T-745 → six phase epics), not labels/milestones
 6. `/whats-next` selects the next batch via `pql ticket list --under <active-phase-epic> --unblocked`
 7. Mutations flow to the git-tracked `.pql/changelog/`; the pre-commit hook exports + stages it
+8. `.pql/changelog/` is write-through data — never `git restore` or hand-edit
+   it; for non-routine planning-store ops (branch moves, recovery) load
+   `Skill(pql)` first
diff --git a/tooling/db/image_connector.py b/tooling/db/image_connector.py
index e6a8d7a81..83a58b41a 100755
--- a/tooling/db/image_connector.py
+++ b/tooling/db/image_connector.py
@@ -2,7 +2,7 @@
 """
 Gemini image generator connector — direct API wrapper.
 
-Generates images via Google's Gemini 2.0 Flash image generation API.
+Generates images via Google's gemini-2.5-flash-image generation API.
 API key from GEMINI_API_KEY env var or config.json.
 
 Usage:
@@ -87,13 +87,13 @@ def generate(prompt, output=None, aspect_ratio="1:1", image_size=None,
     # Add input image if provided (image-to-image)
     if input_image:
         if not os.path.isfile(input_image):
-            print(json.dumps({"ok": False, "error": f"Input image not found: {input_image}"}), indent=2)
+            print(json.dumps({"ok": False, "error": f"Input image not found: {input_image}"}, indent=2))
             sys.exit(1)
         with open(input_image, "rb") as f:
             image_data = base64.b64encode(f.read()).decode("utf-8")
         # Detect mime type
         ext = os.path.splitext(input_image)[1].lower()
-        mime = {"png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg",
+        mime = {".png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg",
                 ".webp": "image/webp"}.get(ext, "image/png")
         parts.append({
             "inlineData": {
@@ -102,20 +102,25 @@ def generate(prompt, output=None, aspect_ratio="1:1", image_size=None,
             }
         })
 
-    # Build enhanced prompt with aspect ratio and size hints
+    # Build enhanced prompt with a size hint. Unlike aspect ratio below, Gemini
+    # has no dedicated resolution parameter for this model — this is a
+    # best-effort prompt hint only and may not be honored.
     enhanced_prompt = prompt
-    if aspect_ratio and aspect_ratio != "1:1":
-        enhanced_prompt += f" Aspect ratio: {aspect_ratio}."
     if image_size:
         enhanced_prompt += f" Resolution: {image_size}."
 
     parts.append({"text": enhanced_prompt})
 
+    generation_config = {"responseModalities": ["TEXT", "IMAGE"]}
+    if aspect_ratio:
+        # Real API parameter (not a prompt hint). Valid values: 1:1, 3:2,
+        # 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.
+        # https://ai.google.dev/gemini-api/docs/image-generation
+        generation_config["imageConfig"] = {"aspectRatio": aspect_ratio}
+
     payload = json.dumps({
         "contents": [{"parts": parts}],
-        "generationConfig": {
-            "responseModalities": ["TEXT", "IMAGE"],
-        }
+        "generationConfig": generation_config
     })
 
     req = urllib.request.Request(
@@ -142,7 +147,7 @@ def generate(prompt, output=None, aspect_ratio="1:1", image_size=None,
         }, indent=2))
         sys.exit(1)
     except Exception as e:
-        print(json.dumps({"ok": False, "error": str(e)}), indent=2)
+        print(json.dumps({"ok": False, "error": str(e)}, indent=2))
         sys.exit(1)
 
     # Extract image data from response