Replace Arrival summary with full LOCKED recipe (validated 2026-03-23)
BetterLeaks / scan (push) Successful in 54s
BetterLeaks / scan (push) Successful in 54s
Expanded from bullet points to complete recipe specification: - Explicit bokeh/blur instructions - Background nudity percentages for B (2-3 topless) and C (20% naked / 30% topless) - 80/20 gender mix - Gently smiling patrons - Result: natural non-intimidating table-appropriate ambiance
This commit is contained in:
@@ -1,12 +1,12 @@
|
||||
---
|
||||
name: npc-nudity-workflow
|
||||
description: Workflow for generating medieval Norse D&D NPC nudity scenes (Undressed/Topless/Underwear/Teasing/Teasing More/Dressed) with Grok Imagine. Use when rendering courtesans/brothel NPCs with explicit nudity constraints, moderation‑avoidance steps, identity consistency, and staged rerender logic (pairs, twins, triples).
|
||||
description: Workflow for generating medieval Norse D&D NPC nudity scenes (Undressed/Topless/Underwear/Teasing/Teasing More/Dressed) with Grok Imagine. Use when rendering courtesans/brothel NPCs with explicit nudity constraints, moderation-avoidance steps, identity consistency, and staged rerender logic (pairs, twins, triples).
|
||||
---
|
||||
|
||||
# NPC Nudity Workflow
|
||||
|
||||
## Overview
|
||||
Use this skill to run the **NPC Nudity Workflow** for Grok‑based NPC renders (courtesans, brothel staff) with clear prompt guardrails, retry logic, and intent‑failure handling.
|
||||
Use this skill to run the **NPC Nudity Workflow** for Grok-based NPC renders (courtesans, brothel staff) with clear prompt guardrails, retry logic, and intent-failure handling.
|
||||
|
||||
**Legal boundary:** All characters rendered using this workflow are **adults aged 18+**. For courtesans specifically, the minimum age is **20+**. This is enforced in prompts (explicit age statements) and character definitions.
|
||||
|
||||
@@ -15,36 +15,45 @@ Use this skill to run the **NPC Nudity Workflow** for Grok‑based NPC renders (
|
||||
- `pass insert x.ai/apikey`
|
||||
- Then scripts read it via `pass show x.ai/apikey | head -n1`.
|
||||
2) Read **references/workflow.md** for the full process, prompts, failure modes, and expected yields.
|
||||
3) Apply the **Default NPC Nudity Workflow** (V4 → V1 retries, unobstructed nipples, per‑person hair matching).
|
||||
3) Apply the **Default NPC Nudity Workflow** (V4 → V1 retries, unobstructed nipples, per-person hair matching).
|
||||
4) **Prevent drift (automatic):** keep required/forbidden phrases in `references/prompt-rules.json` and **call `scripts/prompt_guard.enforce(stage, group, prompt)` inside render scripts** before enqueue. (Use `prompt_lint.py` for audits.)
|
||||
5) **Stage workflow order (rendering):** **Undressed → Dressed (pick keeper) → regenerate Dressed A–E from keeper → Arrival A–C → Teasing → Teasing More → Topless → Underwear.**
|
||||
- **Prereq (Undressed):** before generating **Undressed**, define each courtesan’s **name + extremely specific facial description** (face geometry, nose/brow/jaw/cheekbones/eye shape, unique marks) and **body description detailed enough to draw the same body twice**. Include this anchor in every Undressed prompt to keep identity consistent.
|
||||
5) **Stage workflow order (rendering):** **Undressed → Dressed (pick keeper) → regenerate Dressed A-E from keeper → Arrival A-C → Teasing → Teasing More → Topless → Underwear.**
|
||||
- **Prereq (Undressed):** before generating **Undressed**, define each courtesan's **name + extremely specific facial description** (face geometry, nose/brow/jaw/cheekbones/eye shape, unique marks) and **body description detailed enough to draw the same body twice**. Include this anchor in every Undressed prompt to keep identity consistent.
|
||||
- **Undressed recipe (aligns with old FullyNaked):** use **grok-imagine-image-pro** as primary, apply the **V4 → V1** fallback chain (**V4 twice, then V1 up to 8 times per scene**), and keep **narrative-appropriate nudity wording** (avoid hard explicit-genital phrasing; use short-trimmed pubic hair cues matching head hair **for all genders**).
|
||||
- **Scene selection rule (Undressed):** decide how many scenes you need, then **keep hammering until that count succeeds**. If a scene keeps failing moderation, **switch to a different scene**. **The successful Undressed scenes define the canonical locations** for all later stages.
|
||||
- **QA rule (Undressed, human inspection):** visually inspect all outputs for **intent fails** — images that passed moderation but have anatomical absurdities (e.g., pubic hair extending to knees, impossible proportions). Treat as FAIL and rerender. Keep grinding until all keepers pass basic intent/plausibility checks. This is separate from **moderation fails** (API rejects); intent fails require human judgment.
|
||||
- **Undressed recipe (locked, validated 2026-03-24):** text‑only identity anchors with **short semicolon format** (~25-35 words: hair; eyes; skin; distinctive facial geometry; height; build; cup size). Use **grok‑imagine‑image‑pro**, **V4×2 → V1×2** (4 attempts max), old simple nudity clauses, and **"smiling warmly with inviting, direct eye contact"**. See `references/LOCKED_RECIPE.md` for full specification. Expected fidelity: **75-100%** with good identity design; generic identities fail catastrophically.
|
||||
- **Prereq (Dressed):** before generating **Dressed**, define the **tier‑appropriate clothing description** in text, **detailed enough to render consistently across multiple runs**; use it in Dressed prompts while referencing the chosen Undressed keeper for face/body continuity.
|
||||
- **Arrival (Lobby) stage:** use Dressed as reference, main subject(s) fully dressed; blurred lobby patrons become progressively more nude from A→C for DM-seat ambiance. **Background gender mix:** 80–90% women, 10–20% men. **Keep the crowd sparse and appropriate to room size; do not cram the room (especially Wealthy and up, which should feel spacious).** **If any background patrons wear bottoms at all, they must be small, time‑period‑appropriate bikini‑like bottoms or loincloths (tier‑appropriate); no suspenders/straps or masculine garments. No male patrons with feminine breasts (men must be flat‑chested).**
|
||||
- **QA rule (Undressed, human inspection):** visually inspect all outputs for **intent fails** - images that passed moderation but have anatomical absurdities (e.g., pubic hair extending to knees, impossible proportions). Treat as FAIL and rerender. Keep grinding until all keepers pass basic intent/plausibility checks. This is separate from **moderation fails** (API rejects); intent fails require human judgment.
|
||||
- **Undressed recipe (locked, validated 2026-03-24):** text-only identity anchors with **short semicolon format** (~25-35 words: hair; eyes; skin; distinctive facial geometry; height; build; cup size). Use **grok-imagine-image-pro**, **V4×2 → V1×2** (4 attempts max), old simple nudity clauses, and **"smiling warmly with inviting, direct eye contact"**. See `references/LOCKED_RECIPE.md` for full specification. Expected fidelity: **75-100%** with good identity design; generic identities fail catastrophically.
|
||||
- **Prereq (Dressed):** before generating **Dressed**, define the **tier-appropriate clothing description** in text, **detailed enough to render consistently across multiple runs**; use it in Dressed prompts while referencing the chosen Undressed keeper for face/body continuity.
|
||||
- **Arrival (Lobby) stage (validated recipe 2026-03-23 evening):** use Dressed as reference, main subject(s) fully dressed; blurred lobby patrons become progressively more nude from A→C for DM-seat ambiance.
|
||||
- **Arrival recipe (LOCKED):**
|
||||
- **Model:** `grok-imagine-image` (standard), endpoint `/images/edits`
|
||||
- **Main subject:** fully dressed in exact same outfit as reference; no nudity; eye color must match
|
||||
- **Pose:** standing centered or seated on central furniture (stationary, not walking toward camera)
|
||||
- **Background patrons:** adults only, **out-of-focus and blurred with soft bokeh**; faces not detailed
|
||||
- **Background gender mix:** 80% women, 20% men
|
||||
- **Background expressions:** some gently smiling toward camera/foreground, others chatting with each other
|
||||
- **Crowd density:** sparse and consistent across A-C (do not cram room, especially Comfortable+)
|
||||
- **Lobby progression:**
|
||||
- **A:** Background fully dressed only; sparse crowd
|
||||
- **B:** Small sparse group; only 2-3 topless patrons; no fully naked; rest fully dressed
|
||||
- **C:** Small sparse group (same density as A/B); approximately **20% fully naked, 30% topless, rest fully dressed**
|
||||
- **Arrival A strictness:** background patrons fully clothed with **full chest coverage** (no bare male chests, no bare female breasts, no bikini/lingerie tops, no exposed midriff).
|
||||
- **Future sets:** once a Dressed keeper is chosen for a group, use that single Dressed image as the reference for all Arrival A–C and later stages to lock outfit continuity. **Do not apply retroactively to the current Poor→Wealthy set.**
|
||||
- **Staircase preference (Arrival):** a visible staircase up is canonical, especially Comfortable+. It’s OK to keep **two options per letter** if one includes a staircase.
|
||||
- **Arrival cutoff (current decision):** stop at **A–C** for all tiers; **D/E are removed** due to moderation/artifact issues. Use **standard** model for Arrival.
|
||||
- **Arrival variants:** generate **A1–C6** (18 images per Dressed image; 4–6 are staircase variants).
|
||||
- **Arrival poses:** avoid the “leaning against a beam” look (especially in C). Prefer **standing centered** or **seated on a central furniture piece** that highlights the subject(s).
|
||||
- **Arrival background vibe:** background patrons should appear **gently smiling/curious toward the camera** or **chatting with each other** (still blurred), to avoid bored expressions.
|
||||
- **ArrivalAllSingles overview (all tiers):** generate a lobby “overview” image set showing **all single courtesans together** (dressed, smiling/charming) with the same Arrival A–C rules. **Trigger/precondition:** a Dressed reference is picked for every Single. **Twins:** default is to include **only one** of the twin pair in the overview (space‑saving); leave a configurable option to include both if desired for other users.
|
||||
6) Note **Topless method split (A–B vs C–E)** and **Topless anchor choice (Dressed preferred / Teasing More allowed)** in the workflow.
|
||||
- **Topless batch size (current):** generate **24 Topless images** using letters **A–Z skipping I and O**. Letters **A,B,C,D,E,F,G,H,J,K** use the **A–B method**; letters **L,M,N,P,Q,R,S,T,U,V,W,X,Y,Z** use the **C–E method**. Keep the **A–Z variant mapping consistent across tiers** (use Modest as the template).
|
||||
- **Topless extra variants (ABERX):** generate **A1–A9, B1–B9, E1–E9, R1–R9, X1–X9** using those same per‑letter prompts. Treat the original A/B/E/R/X from the main set as **A0/B0/E0/R0/X0** for continuity.
|
||||
- **Human review step:** while sorting Topless outputs, manually move any images that show extra exposure but are **not fully nude** into a `Topless Plus` folder under the same tier. This bucket is used later for scene‑progression ordering.
|
||||
6) Follow **gate checkpoints**: pause after each A–E batch for human keeper selection, clean filenames, and ask which **table-facing** tier to expose: **Underwear L** (large loincloth‑type cover), **Underwear S** (small modern‑size cover), or **Undressed** (actually naked). (Undressed is still generated regardless.)
|
||||
7) **Before Teasing More**: inspect Teasing keepers and record which shoulder is off (viewer‑left/right/both). Do this **separately for singles vs doubles/triples**, and for doubles/triples record **per‑image, per‑person** shoulder directions (each scene can differ). Include explicit shoulder directions in Teasing More prompts (per person and per pair/triple).
|
||||
- **Staircase rule:** variants 1-3 no stairs; variants 4-6 include visible staircase going up.
|
||||
- **Arrival variants:** generate **A1–C6** (18 images per Dressed image).
|
||||
- **Result quality (validated):** B and C produce natural, non-intimidating table-appropriate ambiance with properly blurred nude/topless background patrons.
|
||||
- **ArrivalAllSingles overview (all tiers):** generate a lobby "overview" image set showing **all single courtesans together** (dressed, smiling/charming) with the same Arrival A-C rules. **Trigger/precondition:** a Dressed reference is picked for every Single. **Twins:** default is to include **only one** of the twin pair in the overview (space-saving); leave a configurable option to include both if desired for other users.
|
||||
6) Note **Topless method split (A-B vs C-E)** and **Topless anchor choice (Dressed preferred / Teasing More allowed)** in the workflow.
|
||||
- **Topless batch size (current):** generate **24 Topless images** using letters **A-Z skipping I and O**. Letters **A,B,C,D,E,F,G,H,J,K** use the **A-B method**; letters **L,M,N,P,Q,R,S,T,U,V,W,X,Y,Z** use the **C-E method**. Keep the **A-Z variant mapping consistent across tiers** (use Modest as the template).
|
||||
- **Topless extra variants (ABERX):** generate **A1-A9, B1-B9, E1-E9, R1-R9, X1-X9** using those same per-letter prompts. Treat the original A/B/E/R/X from the main set as **A0/B0/E0/R0/X0** for continuity.
|
||||
- **Human review step:** while sorting Topless outputs, manually move any images that show extra exposure but are **not fully nude** into a `Topless Plus` folder under the same tier. This bucket is used later for scene-progression ordering.
|
||||
6) Follow **gate checkpoints**: pause after each A-E batch for human keeper selection, clean filenames, and ask which **table-facing** tier to expose: **Underwear L** (large loincloth-type cover), **Underwear S** (small modern-size cover), or **Undressed** (actually naked). (Undressed is still generated regardless.)
|
||||
7) **Before Teasing More**: inspect Teasing keepers and record which shoulder is off (viewer-left/right/both). Do this **separately for singles vs doubles/triples**, and for doubles/triples record **per-image, per-person** shoulder directions (each scene can differ). Include explicit shoulder directions in Teasing More prompts (per person and per pair/triple).
|
||||
6.1) **Required step:** use a subagent with model **`openrouter/x-ai/grok-4.1-fast`** to inspect all Teasing images and write a shoulder map JSON (e.g., `/home/rick/.openclaw/workspace/tmp/<tier>_teasing_shoulder_map.json`). This is mandatory before all TeasingMore runs.
|
||||
7) **Underwear folder split (table‑facing):** Underwear L=A–C/plain, Underwear S=D–F (g‑string). Apply from Modest onward.
|
||||
7) **Underwear folder split (table-facing):** Underwear L=A-C/plain, Underwear S=D-F (g-string). Apply from Modest onward.
|
||||
7.1) **Boundary Test (explicit undressed):** prompt used:
|
||||
“Totally nude female, explicit genitalia and nipples visible, zero fabric or censorship, bare vulva in full view, reclining pose, high detail erotic D&D fantasy.”
|
||||
Run **A–G** variants and **try each single up to 7 moderation rejects** before fail. (Singles only.)
|
||||
7.2) **Teasing QA (stage only):** Singles reject unless **exactly 1** nipple visible. Doubles/DoubleTwins reject if **0, 3, or 4** nipples visible (only **1–2** allowed). Triples reject if **>4** nipples visible.
|
||||
"Totally nude female, explicit genitalia and nipples visible, zero fabric or censorship, bare vulva in full view, reclining pose, high detail erotic D&D fantasy."
|
||||
Run **A-G** variants and **try each single up to 7 moderation rejects** before fail. (Singles only.)
|
||||
7.2) **Teasing QA (stage only):** Singles reject unless **exactly 1** nipple visible. Doubles/DoubleTwins reject if **0, 3, or 4** nipples visible (only **1-2** allowed). Triples reject if **>4** nipples visible.
|
||||
8) **Concurrency default:** use **25 workers** for `grok-imagine-image` (standard). `grok-imagine-image-pro` is limited (~30 images/min), so keep concurrency lower as needed.
|
||||
8) Use the provided scripts (if bundled) to rerender placeholders and manage staging/logging.
|
||||
|
||||
@@ -56,22 +65,22 @@ Prerequisite: The human user has selected favorites across all dress stages and
|
||||
|
||||
**ArrivalAll packaging (tier root):** place the selected **ArrivalAll** overview images **directly in the tier folder** (same level as passport photos, not inside Per Courtesan). When tier has both genders, generate three sets and name packaged finals: `Arrival (Men, Lobby type A/B/C).jpg`, `Arrival (Ladies, Lobby type A/B/C).jpg`, `Arrival (All, Lobby type A/B/C).jpg`. Add `with stairs` suffix if applicable. The raw composites are generated as `ArrivalAllMen_A1-C6`, `ArrivalAllLadies_A1-C6`, `ArrivalAllSingles_A1-C6` (18 images each); human selects the best per letter.
|
||||
|
||||
**Passport lineup thumbnails (tier root):** generate **`01-Name.jpg`** etc from each Single’s **Dressed.jpg** so the lineup is visible at a glance (no counting in ArrivalAllSingles). Use **OpenCV face detection** on the **upper half** of the Dressed image and crop a square centered on the **largest face** with ~2× padding. **Human review required**: if any face is poorly framed, regenerate that individual using **Pillow fallback** (upper‑half, center‑X square crop). **Do not generate** `01-Name.txt` markers anymore; the images fully replace them. Dependencies: OpenCV in venv `~/.openclaw/workspace/.venv-opencv` (opencv‑python + numpy). Pillow fallback in `~/.openclaw/workspace/.venv-pillow`.
|
||||
**Passport lineup thumbnails (tier root):** generate **`01-Name.jpg`** etc from each Single's **Dressed.jpg** so the lineup is visible at a glance (no counting in ArrivalAllSingles). Use **OpenCV face detection** on the **upper half** of the Dressed image and crop a square centered on the **largest face** with ~2× padding. **Human review required**: if any face is poorly framed, regenerate that individual using **Pillow fallback** (upper-half, center-X square crop). **Do not generate** `01-Name.txt` markers anymore; the images fully replace them. Dependencies: OpenCV in venv `~/.openclaw/workspace/.venv-opencv` (opencv-python + numpy). Pillow fallback in `~/.openclaw/workspace/.venv-pillow`.
|
||||
|
||||
**Post-packaging cleanup:** create a folder named **"Stages and Raw Files"** under the tier, and move **all folders at the tier root except "Per Courtesan"** into it (Arrival/Dressed/Teasing/Teasing More/Topless/Topless Plus/Underwear/Undressed/etc.) to reduce DM-seat noise.
|
||||
**Re-run packaging note:** if **"Stages and Raw Files"** already exists, treat it as the **stage root** (read inputs from there) rather than the tier root.
|
||||
|
||||
**Multi-variant rule (Topless / Topless Plus / Undressed):** these stages may have multiple files per courtesan scene. If the filenames contain bracketed indices like `[1]`, `[2]`, `[3]` (or any bracket/brace/paren variants such as `{3}` or `(3)`), then copy them as `Topless 1.jpg`, `Topless 2.jpg`, etc. (same for Topless Plus / Undressed). **If both bracketed and unnumbered files exist for the same stage, halt with an error** (the human needs to resolve the ambiguity first). The folder name remains **"Per Courtesan"**.
|
||||
|
||||
**Preflight requirement (packaging):** before packaging, **fail fast** if any non‑Arrival stage files still include stage letters (e.g., `_A__`, `_B__`, `_A_E__`, or `TeasingMoreC__`). These must be stripped before packaging.
|
||||
**Preflight requirement (packaging):** before packaging, **fail fast** if any non-Arrival stage files still include stage letters (e.g., `_A__`, `_B__`, `_A_E__`, or `TeasingMoreC__`). These must be stripped before packaging.
|
||||
|
||||
## References
|
||||
- **references/workflow.md** — the full process notes, prompt clauses, failure modes, yields, and timing.
|
||||
- **references/workflow.md** - the full process notes, prompt clauses, failure modes, yields, and timing.
|
||||
|
||||
## Resources
|
||||
### references/
|
||||
- **workflow.md** — detailed operational guide and prompt templates.
|
||||
- **prompt-rules.json** — required/forbidden phrases by stage/group (single source of truth).
|
||||
- **workflow.md** - detailed operational guide and prompt templates.
|
||||
- **prompt-rules.json** - required/forbidden phrases by stage/group (single source of truth).
|
||||
### scripts/
|
||||
- **prompt_lint.py** — lint rendered prompts against prompt‑rules.json (batch audit).
|
||||
- **prompt_guard.py** — programmatic guard for automatic lint (call `enforce(...)` in scripts).
|
||||
- **prompt_lint.py** - lint rendered prompts against prompt-rules.json (batch audit).
|
||||
- **prompt_guard.py** - programmatic guard for automatic lint (call `enforce(...)` in scripts).
|
||||
|
||||
Reference in New Issue
Block a user