@lemoncat7/dsh-partner 2.5.0 → 2.7.0

package/resources/skills/director-skills/upstream/SKILL.md

@@ -0,0 +1,114 @@
+---
+name: director-skills
+description: "Master router for a modular AI filmmaking skill suite. Use for story development, screenplay scenes, shot lists, visual bibles, cinematic image/video prompts, continuity, prompt diagnostics, and model-specific exports when the agent should choose and orchestrate the right director-skills sub-workflows."
+---
+
+# director-skills
+
+Use this as the master conductor for the `director-skills` suite. It is intentionally a router, not a replacement for the specialized skills under `skills/`.
+
+The suite can operate in two modes:
+
+- **Native modular mode**: the host agent registers every folder under `skills/` as its own skill.
+- **Single uploaded skill mode**: the host agent loads this root `SKILL.md`, then routes internally to the relevant sub-skill files.
+
+## Prime Directive
+
+Do not treat the README as the full operating procedure. Use it only for orientation. For real work, select the relevant sub-skill, read that skill's `SKILL.md`, and load supporting files only as needed.
+
+## When To Use
+
+Use this master skill when:
+
+- The user asks for help with AI filmmaking, cinematic storytelling, image prompting, video prompting, or prompt packaging.
+- The user uploads or links the whole `director-skills` repository and says to learn, use, import, or follow it.
+- The task spans multiple production stages, such as idea to screenplay to shot list to prompts.
+- The user asks which director-skills workflow applies.
+- The agent environment supports only one uploaded skill or knowledge bundle.
+- The request is broad enough that routing is needed before execution.
+
+## When Not To Use
+
+Do not use this skill for:
+
+- Tasks unrelated to filmmaking, story, cinematic prompting, continuity, visual development, or model-specific prompt export.
+- Live model capability claims that require current official documentation unless the user provides sources or the agent can verify them.
+- Requests to bypass safety systems, imitate living artists in a deceptive way, expose private files, or reproduce copyrighted source guides.
+
+If a host system has already activated a specific sub-skill that fully covers the request, follow that sub-skill directly. If model naming, continuity, or failure diagnosis appears, add the appropriate supporting sub-skill.
+
+## Core Method
+
+Every task moves through three layers:
+
+1. **Creative intent**: story, theme, character, emotion, scene purpose, dramatic action, genre, tone, and audience experience.
+2. **Cinematic execution**: shot type, framing, angle, camera movement, lens, lighting, blocking, production design, texture, realism, and visual continuity.
+3. **Model adaptation**: prompt structure, syntax, length, parameters, negative support, references, aspect ratio, duration, resolution, seeds, and known model failure modes.
+
+Do not assume one prompt format works across all image and video models.
+
+## Routing Protocol
+
+1. Classify the user's task by production stage and requested output.
+2. Choose one primary sub-skill and zero to three supporting sub-skills.
+3. Read the primary `skills/<skill-name>/SKILL.md`.
+4. Read supporting sub-skill `SKILL.md` files only when their decision logic is needed.
+5. Load references, templates, checklists, examples, or tests only when the selected sub-skill asks for them or the task requires the specific artifact.
+6. Produce the requested output and include the quality checks from the active sub-skill.
+7. If the task crosses stages, hand off cleanly by naming the next recommended sub-skill and preserving continuity anchors.
+
+## Sub-Skill Router
+
+| User intent | Primary sub-skill | Add these when relevant |
+|---|---|---|
+| Raw idea, premise, logline, treatment, outline, short-story plan | `short-film-development` | `style-cinematography-director`, `character-location-prop-bible` |
+| Scene writing, dialogue, beats, pacing, Fountain, visual action | `screenplay-and-scene-writing` | `short-film-development`, `character-location-prop-bible` |
+| Script to shot list, storyboard plan, camera plan, scene breakdown, asset list | `shotlist-and-visual-breakdown` | `style-cinematography-director`, `character-location-prop-bible` |
+| Character sheet, location reference, prop prompt, poster, keyframe, cinematic still | `cinematic-image-prompting` | `character-location-prop-bible`, `style-cinematography-director`, `model-adaptation` |
+| Text-to-video, image-to-video, motion prompt, time blocks, camera movement, audio cues | `cinematic-video-prompting` | `shotlist-and-visual-breakdown`, `model-adaptation`, `prompt-iteration-and-diagnostics` |
+| Continuity bible, identity anchors, wardrobe, prop state, location rules, style bible | `character-location-prop-bible` | `cinematic-image-prompting`, `shotlist-and-visual-breakdown` |
+| Genre look, cinematography, lensing, lighting, realism, visual language | `style-cinematography-director` | `shotlist-and-visual-breakdown`, `cinematic-image-prompting`, `cinematic-video-prompting` |
+| Failed output, drift, bad motion, wrong face, weak realism, revision loop | `prompt-iteration-and-diagnostics` | Relevant image/video skill, `model-adaptation`, `character-location-prop-bible` |
+| Named model, prompt conversion, multi-tool export, settings, capability mismatch | `model-adaptation` | Relevant image/video/story skill |
+
+## Model Trigger Rule
+
+If the user names an image or video model, always consider `model-adaptation`.
+
+Examples include Midjourney, DALL-E, GPT-image, Stable Diffusion, FLUX, Ideogram, Grok/Aurora, Runway, Pika, Luma, Kling, Veo, Sora, Hailuo, Seedance, Gemini image, Nano Banana, Le Chat, or any newly named model. If the suite has no clear profile for the tool, mark capabilities as `unknown` instead of guessing.
+
+## Multi-Skill Pipelines
+
+For end-to-end work, use this order unless the user asks for a narrower artifact:
+
+```text
+short-film-development
+-> screenplay-and-scene-writing
+-> character-location-prop-bible
+-> style-cinematography-director
+-> shotlist-and-visual-breakdown
+-> cinematic-image-prompting
+-> cinematic-video-prompting
+-> model-adaptation
+-> prompt-iteration-and-diagnostics
+```
+
+Keep the output moving. If the user asks for an artifact, produce the artifact; do not stop at a plan unless they asked for a plan.
+
+## Output Discipline
+
+Before finalizing, verify:
+
+- The selected sub-skill matches the user's actual requested output.
+- Creative intent, cinematic execution, and model adaptation are not collapsed into one vague prompt.
+- Continuity anchors are preserved when characters, locations, costumes, props, or style recur.
+- Model-specific claims come from the model-adaptation references, user-provided docs, or are marked `unknown`.
+- The output is immediately usable: a scene, list, bible entry, prompt, export package, diagnosis, or next-step handoff.
+
+## Supporting Files
+
+- For full user-facing orientation, read `README.md`.
+- For the public-safe research basis, read `RESEARCH_PROVENANCE.md`.
+- For complete file inventory, read `MANIFEST.md`.
+- For operational work, route into `skills/<skill-name>/SKILL.md` and that skill's own folders.
+

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/SKILL.md

@@ -0,0 +1,81 @@
+---
+name: character-location-prop-bible
+description: "Create and maintain continuity bibles for characters, locations, props, costumes, color palettes, style rules, reference assets, and cross-shot visual anchors."
+---
+
+# character-location-prop-bible
+
+## When to use
+- The user needs consistency across images, shots, scenes, or models.
+- The user asks for a character bible, location bible, prop list, wardrobe continuity, style bible, or continuity checklist.
+- The project is moving from idea/script into repeatable image/video generation.
+
+## When not to use
+- The user only wants one disposable prompt with no continuity need.
+- The task is pure model parameter research.
+- The user asks for story structure without visual asset continuity.
+
+## Required inputs
+- Project title or scene
+- Existing descriptions or outputs
+- Characters/locations/props/costumes to track
+- Target visual style
+
+## Optional inputs
+- Reference images
+- Shot list
+- Model-specific requirements
+- Palette or genre anchors
+- Version history
+
+## Workflow
+1. Create stable IDs for characters, locations, props, costumes, and style rules.
+2. Record immutable anchors: identity, silhouette, proportions, palette, materials, prop condition, location geography, and lighting baseline.
+3. Separate mutable state: emotion, wardrobe changes, damage, weather, time of day, prop position, and story-state changes.
+4. Create golden-image reference prompt entries for clean sheets before dramatic variants.
+5. After each generated output or script change, update only the changed fields and preserve a changelog note.
+6. Prepare a continuity packet for prompting: minimal anchors for each shot plus exclusions for drift-prone details.
+
+## Decision logic
+- If a detail must never change, place it in immutable anchors.
+- If a detail changes by scene, place it in state timeline.
+- If two sources conflict, prefer the latest user-approved bible entry and log the conflict.
+- If model transfer causes drift, route to model-adaptation and use reference-image workflows where supported.
+
+## Output formats
+- Character profile
+- Location profile
+- Prop profile
+- Costume profile
+- Visual style bible
+- Continuity checklist
+- Reference asset index
+- Golden image plan
+- State timeline
+
+## Quality checks
+- Every recurring asset has an ID and stable anchor fields.
+- Mutable state is tied to scene/shot numbers.
+- Prompt anchors are concise enough to reuse.
+- Style rules distinguish creative intent from execution details.
+- Conflicts are logged instead of silently resolved.
+
+## Anti-patterns
+- Changing character descriptions between prompts
+- Mixing immutable identity with temporary emotion
+- Overlong bible entries that agents cannot practically reuse
+- Ignoring prop state after action beats
+- Relying on text-only consistency when references are available.
+
+## Exit criteria
+- A concise continuity packet exists for downstream screenplay, shot list, image prompts, video prompts, and model exports.
+
+## Supporting files
+Read only the supporting file needed for the active task:
+- `references/continuity_system.md`
+- `references/visual_bible_system.md`
+- `references/identity_anchors.md`
+- `references/golden_image_workflow.md`
+- `templates/character_profile.md`
+- `templates/location_profile.md`
+- `templates/continuity_checklist.md`

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/checklists/continuity_checklist.md

@@ -0,0 +1,8 @@
+# Continuity Checklist
+
+- Recurring assets have IDs.
+- Immutable anchors are separated from mutable state.
+- Scene/shot state changes are logged.
+- Reference assets are named.
+- Prompt anchor blocks are concise.
+- Conflicts are recorded and resolved explicitly.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/examples/character_continuity_update.md

@@ -0,0 +1,9 @@
+# Example: Character Continuity Update
+
+## Change
+After Scene 2, the guard's left sleeve is wet from the leaking basement pipe.
+
+## Bible update
+- COST-001 mutable state: left cuff damp, darker fabric edge, begins Scene 3 and persists through Scene 4.
+- Prompt anchor: navy guard jacket with damp left cuff, subtle water-darkened fabric, otherwise unchanged wardrobe.
+- Conflict note: do not change jacket color; only dampness changes.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/examples/location_continuity_update.md

@@ -0,0 +1,9 @@
+# Example: Location Continuity Update
+
+## Change
+Door 12B is opened in Shot S01-004.
+
+## Bible update
+- LOC-001 basement hallway state: storage door 12B ajar after S01-004; dust disturbed at threshold.
+- Lighting state: corridor remains green fluorescent, room interior unlit except flashlight beam.
+- Prompt anchor: chipped green metal door marked 12B, now open six inches, dust streak at threshold.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/references/continuity_system.md

@@ -0,0 +1,21 @@
+# Continuity System
+
+## Asset IDs
+Use stable IDs:
+- CH-001 character
+- LOC-001 location
+- PROP-001 prop
+- COST-001 costume
+- STYLE-001 visual rule
+
+## Immutable vs mutable
+Immutable anchors stay stable across the project: face shape, age range, height impression, silhouette, location geography, prop design, palette rules.
+
+Mutable state changes by scene/shot: emotion, dirt, damage, weather, lighting state, prop position, wardrobe layer, injury, time of day.
+
+## Update protocol
+1. Record source of change.
+2. Update only changed fields.
+3. Add scene/shot state row.
+4. Note conflicts and chosen resolution.
+5. Generate a concise prompt anchor block.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/references/golden_image_workflow.md

@@ -0,0 +1,29 @@
+# Golden Image Workflow
+
+Golden images are approved still references used to keep generated shots visually consistent.
+
+## When to create golden images
+
+- Before image-to-video generation.
+- Before multi-shot sequences with recurring characters.
+- Before a model transfer where identity/style drift is likely.
+- Before marketing assets if the look is not fully locked.
+
+## What to lock
+
+- Characters: neutral face, full-body silhouette, wardrobe, hair, skin, signature prop.
+- Locations: readable geography, landmarks, time-of-day baseline, palette, material texture.
+- Props: scale, material, wear, markings, state variants.
+- Style: palette, lighting quality, grain/finish, realism level.
+
+## Production use
+
+1. Generate 2-4 candidate reference stills.
+2. Select one approved image per asset or state.
+3. Assign asset IDs and store prompt/settings/source model.
+4. Use the still as I2V start frame or reference asset where supported.
+5. Generate 2-3 video variations per shot and discard outputs with identity, physics, or morphology failures.
+
+## Quality rule
+
+Do not use a dramatic, shadow-heavy, occluded still as the only identity reference. Create a clean reference first, then dramatic variants.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/references/identity_anchors.md

@@ -0,0 +1,14 @@
+# Identity Anchors
+
+## Character anchor fields
+- Name/ID.
+- Age range and presence.
+- Face shape, hair, eyes, skin details.
+- Silhouette and posture.
+- Wardrobe baseline and materials.
+- Distinctive prop or gesture.
+- Emotional baseline.
+- Forbidden drift: changes that would break identity.
+
+## Reference sheet guidance
+Generate neutral, well-lit, front/three-quarter/profile views before dramatic variants. Avoid extreme expressions, heavy shadows, or occluding props in the reference sheet.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/references/visual_bible_system.md

@@ -0,0 +1,15 @@
+# Visual Bible System
+
+A visual bible turns creative direction into reusable prompt anchors.
+
+## Sections
+- Project style premise.
+- Palette rules: dominant, accent, forbidden colors.
+- Camera rules: preferred shot scale, movement, lens distance.
+- Lighting rules: source, contrast, color temperature.
+- Texture rules: skin, fabric, surfaces, atmosphere.
+- Design rules: props, architecture, costume language.
+- Model notes: reference images, seeds, aspect ratios, known failure risks.
+
+## Prompt packet
+For each shot, include only the anchors that matter. Avoid pasting the entire bible into every prompt.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/templates/character_profile.md

@@ -0,0 +1,15 @@
+# Character Profile Template
+
+- ID:
+- Name:
+- Role in story:
+- Age range/presence:
+- Face/hair/eyes/skin anchors:
+- Silhouette/posture/gait:
+- Wardrobe baseline:
+- Distinctive prop/gesture:
+- Emotional baseline:
+- Mutable states by scene:
+- Forbidden drift:
+- Reference assets:
+- Prompt anchor block:

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/templates/continuity_checklist.md

@@ -0,0 +1,11 @@
+# Continuity Checklist Template
+
+Before generating a shot, confirm:
+- Character ID and current emotional/physical state.
+- Wardrobe state.
+- Prop state and position.
+- Location/time/weather state.
+- Screen direction and blocking.
+- Palette/lighting continuity.
+- Reference assets attached or named.
+- Model-specific consistency method selected.

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/templates/costume_profile.md

@@ -0,0 +1,12 @@
+# Costume Profile Template
+
+- ID:
+- Character:
+- Garment pieces:
+- Materials/textures:
+- Palette:
+- Fit/silhouette:
+- Wear/damage/dirt states:
+- Scene continuity:
+- Forbidden drift:
+- Prompt anchor block:

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/templates/location_profile.md

@@ -0,0 +1,14 @@
+# Location Profile Template
+
+- ID:
+- Name:
+- Story function:
+- Geography/layout:
+- Key landmarks:
+- Lighting baseline:
+- Palette/materials:
+- Weather/time variants:
+- Props/set dressing:
+- Mutable states by scene:
+- Reference assets:
+- Prompt anchor block:

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/templates/prop_profile.md

@@ -0,0 +1,13 @@
+# Prop Profile Template
+
+- ID:
+- Name:
+- Story function:
+- Physical description:
+- Material/texture:
+- Initial state:
+- State changes:
+- Who handles it:
+- Continuity risks:
+- Reference assets:
+- Prompt anchor block:

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/templates/visual_style_bible.md

@@ -0,0 +1,15 @@
+# Visual Style Bible Template
+
+- Project:
+- Audience experience:
+- Genre grammar:
+- Palette:
+- Camera rules:
+- Lens rules:
+- Movement rules:
+- Lighting rules:
+- Texture/realism rules:
+- Production design rules:
+- Audio tone:
+- Model notes:
+- Forbidden style drift:

package/resources/skills/director-skills/upstream/skills/character-location-prop-bible/tests/continuity_tests.md

@@ -0,0 +1,6 @@
+# Continuity Tests
+
+- Every recurring asset receives a stable ID.
+- Immutable anchors are not overwritten by mutable scene states.
+- Conflicts are logged.
+- Prompt anchor block stays concise.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: cinematic-image-prompting
+description: "Generate cinematic image prompts for characters, locations, props, costumes, keyframes, posters, concept art, reference sheets, and scene stills using creative intent plus camera, lighting, texture, and continuity anchors."
+---
+
+# cinematic-image-prompting
+
+## When to use
+- The user wants still image prompts, character sheets, location references, prop/costume concepts, keyframes, posters, thumbnails, concept art, or scene stills.
+- The user needs image prompts before video generation to lock identity, composition, or style.
+- The task requires photorealism, cinematic language, texture, camera/lens terms, or negative/positive constraints.
+
+## When not to use
+- The user asks for moving video prompts only.
+- The user asks to adapt a finished prompt to a named model only.
+- The user wants actual image generation rather than prompt writing, unless paired with image generation tools.
+
+## Required inputs
+- Subject or asset type
+- Purpose of image
+- Creative intent
+- Continuity anchors
+- Style/genre/tone
+
+## Optional inputs
+- Target model
+- Aspect ratio
+- Reference images
+- Text/logo needs
+- Photorealism level
+- Negative constraints
+
+## Workflow
+1. Start with creative intent: who/what, why this image exists, emotional read, and audience impression.
+2. Build cinematic execution: subject specificity, action/pose, environment, framing, lens feel, lighting direction/quality, color, texture, and realism level.
+3. Add continuity anchors from the bible: face, wardrobe, prop state, location geography, palette, and design rules.
+4. Select format: natural prose, priority stack, shot card, or model-specific export through model-adaptation.
+5. Handle exclusions according to model behavior: true negative field where supported, positive restatement where not.
+6. Create variants only by changing one meaningful variable at a time.
+7. Add quality checklist and revision hooks.
+
+## Decision logic
+- If character consistency matters, create a clean reference sheet before dramatic scene stills.
+- If the image contains text, route to model-adaptation and choose a model/profile that supports text better.
+- If the style is too broad, ask for or infer one strong visual anchor plus specific execution details.
+- If the target model is unknown, output a universal prompt brief and flag model-dependent fields.
+
+## Output formats
+- Universal image prompt brief
+- Cinematic still prompt
+- Character/location/prop reference prompt
+- Negative or positive constraint block
+- Prompt variants
+- Model-adaptation handoff
+
+## Quality checks
+- Prompt has a clear subject, action/pose, environment, composition, lighting, style, texture, and output purpose.
+- The first sentence front-loads the most important identity and scene elements.
+- No contradictory style or lighting instructions.
+- Continuity anchors are explicit and not overlong.
+- Model-specific claims are not guessed.
+
+## Anti-patterns
+- Keyword soup for natural-language models
+- Overloaded prompts with many competing references
+- Negative prompts in models that lack negative support
+- Vague words like epic or cinematic without camera/lighting specifics
+- Using copyrighted or trademarked specifics when a generic visual description is safer.
+
+## Exit criteria
+- The prompt is ready for universal use or handoff to model-adaptation for a named image model.
+
+## Supporting files
+Read only the supporting file needed for the active task:
+- `references/image_prompt_components.md`
+- `references/lighting_and_photorealism.md`
+- `references/prompt_anatomy.md`
+- `references/negative_prompt_patterns.md`
+- `templates/cinematic_still_prompt.md`
+- `templates/character_reference_sheet_prompt.md`

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/checklists/image_prompt_quality_checklist.md

@@ -0,0 +1,10 @@
+# Image Prompt Quality Checklist
+
+- Purpose and asset type are clear.
+- Subject identity is specific.
+- Action/pose is visible.
+- Environment is grounded.
+- Framing/lens/light/palette are specified.
+- Texture/realism details prevent generic output.
+- Continuity anchors are concise.
+- Model-dependent negatives are handled correctly.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/examples/shot_list_to_image_prompts.md

@@ -0,0 +1,4 @@
+# Example: Shot List to Image Prompts
+
+## Shot S01-002 keyframe
+Extreme close-up of a worn brass key stamped "12B" shaking between the night guard's fingers, chipped green basement paint blurred behind it, flashlight edge light grazing the metal, shallow macro focus, visible fingerprints and scratches, tense quiet realism, cool fluorescent color cast, 16:9.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/references/image_prompt_components.md

@@ -0,0 +1,22 @@
+# Image Prompt Components
+
+## Universal image stack
+1. Purpose: character sheet, scene still, keyframe, poster, prop, location, product, concept art.
+2. Subject: stable identity, silhouette, age range, wardrobe/materials, pose or action.
+3. Context: location, time, weather, story moment, production design.
+4. Composition: shot scale, angle, framing, aspect ratio, negative space.
+5. Camera/lens: focal length feel, depth of field, sensor/format only if useful.
+6. Lighting: source, direction, quality, contrast, color temperature.
+7. Palette/style: genre, color rule, medium, realism level.
+8. Texture: skin, fabric, surface, atmospheric details.
+9. Continuity: bible IDs, prop state, wardrobe, location state.
+10. Constraints: model-appropriate negatives or positive restatements.
+
+## Prompt formats
+- Natural prose: best for LLM-mediated models.
+- Priority stack: best when front-loading and clarity matter.
+- Shot card: concise, useful for production packages.
+- Model-specific: route through model-adaptation.
+
+## Completeness checklist
+A prompt is complete when it answers: what is seen, why it matters, where it is, how it is framed, how it is lit, what texture makes it real, and what must stay consistent.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/references/lighting_and_photorealism.md

@@ -0,0 +1,19 @@
+# Lighting and Photorealism
+
+Photorealism improves when prompts describe physical light and material response instead of generic quality words.
+
+## Lighting controls
+- Direction: camera-left, camera-right, overhead, backlight, underlight, window-side.
+- Quality: soft, hard, diffused, specular, bounced, practical, fluorescent, sodium vapor, neon.
+- Source: window, candle, monitor, streetlight, moon, overcast sky, softbox, bare bulb.
+- Contrast: high-key, low-key, chiaroscuro, silhouette, rim light, fill ratio.
+- Color temperature: cool blue dawn, warm tungsten, sickly green fluorescent, mixed neon.
+
+## Realism anchors
+- Skin: pores, fine lines, natural asymmetry, subsurface warmth, no waxy smoothing.
+- Fabric: weave, wrinkles, seams, damp edges, lint, worn cuffs.
+- Environment: dust, condensation, scratches, fingerprints, water streaks, chipped paint.
+- Optics: shallow depth of field, halation, soft highlight bloom, film grain, slight lens breathing.
+
+## Anti-plastic guardrails
+Use concise constraints: natural skin texture, imperfect fabric, physical shadows, grounded contact, realistic reflections, no waxy CGI surface.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/references/negative_prompt_patterns.md

@@ -0,0 +1,18 @@
+# Negative Prompt Patterns
+
+Do not assume negative prompts work in every model.
+
+## If the model supports a negative field
+Use only the few failures that would ruin the output. Prefer concrete artifacts: extra fingers, warped hands, text artifacts, logo distortion, identity drift, blurry face, morphing background.
+
+## If the model does not support negatives
+Restate the desired positive condition:
+- Instead of "no blur": sharp focus on the subject's eyes.
+- Instead of "no extra people": single person alone in the frame.
+- Instead of "no morphing": stable face shape and unchanged wardrobe throughout.
+
+## Video constraint cap
+For video, limit constraints to the top 2 to 4 risks. Over-constraining can create prompt noise.
+
+## Safety/IP hygiene
+Do not provide bypass tactics. Replace named protected characters/logos with visual descriptions unless the user has a legitimate, allowed use case and the model supports it.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/references/prompt_anatomy.md

@@ -0,0 +1,18 @@
+# Prompt Anatomy
+
+Use the three-layer model.
+
+## Layer 1: Creative intent
+Story purpose, character emotion, theme pressure, audience experience, genre, and scene function.
+
+## Layer 2: Cinematic execution
+Shot type, framing, camera angle, lens, lighting, blocking, production design, texture, realism level, audio if video, and continuity anchors.
+
+## Layer 3: Model-specific adaptation
+Prompt syntax, length, parameters, negative support, reference handling, aspect ratio, duration, resolution, seed/reference controls, known failure modes.
+
+## Front-loading rule
+Place the identity, action, setting, and desired medium in the opening sentence. Put secondary style and constraints later.
+
+## Prompt economy
+Remove filler such as "we see" or repeated adjectives. Keep every clause tied to a visible or audible result.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/templates/character_reference_sheet_prompt.md

@@ -0,0 +1,5 @@
+# Character Reference Sheet Prompt Template
+
+Create a clean cinematic character reference sheet for [CH-ID/name]. [Age range], [face/hair/eyes/skin anchors], [silhouette/posture], wearing [wardrobe baseline with material and color]. Neutral expression plus subtle variants, front view, three-quarter view, profile, full-body view. Even soft studio lighting, plain neutral background, natural skin texture, accurate proportions, consistent features across all views. Purpose: lock identity for later image/video prompts. Aspect ratio: [ratio].
+
+Constraints: single character only, no dramatic shadows, no occluding props, no text unless the target model supports it.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/templates/cinematic_still_prompt.md

@@ -0,0 +1,3 @@
+# Cinematic Still Prompt Template
+
+[Subject/character ID] [visible action or pose] in [specific location], at [story moment/emotional state]. Framing: [shot size, angle, composition]. Lens/camera feel: [focal length/depth/format]. Lighting: [source, direction, quality, contrast, color temperature]. Production design: [props, wardrobe, set dressing]. Texture/realism: [skin/fabric/surfaces/atmosphere]. Palette/style: [genre, grade, realism level]. Continuity: [anchors]. Aspect ratio: [ratio].

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/templates/keyframe_prompt.md

@@ -0,0 +1,6 @@
+# Keyframe Prompt Template
+
+Purpose: first frame / last frame / transition keyframe.
+
+Prompt:
+[Character/subject] at the exact [start/end] state of the action: [pose, gaze, prop state, location state]. Composition matches shot [ID]: [framing, lens feel, camera angle]. Lighting and palette remain [continuity rule]. This keyframe should be clean, readable, and animation-ready for [target video model/workflow].

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/templates/location_reference_prompt.md

@@ -0,0 +1,3 @@
+# Location Reference Prompt Template
+
+Create a location reference image for [LOC-ID/name], [specific geography and architecture], at [time/weather]. Show the layout clearly with foreground, midground, and background landmarks. Lighting: [source/direction/quality]. Palette: [dominant/accent colors]. Texture: [surfaces, wear, atmosphere]. Camera: [wide/medium/lens feel]. Purpose: establish continuity for later shots.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/templates/prop_prompt.md

@@ -0,0 +1,3 @@
+# Prop Prompt Template
+
+Create a prop reference image for [PROP-ID/name], [object type, size, materials, condition, distinctive marks]. Present on a simple surface with [lighting] so silhouette, texture, and important details are readable. Include [state variants if needed]. No hands unless scale is essential. Purpose: continuity anchor.

package/resources/skills/director-skills/upstream/skills/cinematic-image-prompting/tests/image_prompt_completeness_tests.md

@@ -0,0 +1,5 @@
+# Image Prompt Completeness Tests
+
+- Prompt includes purpose, subject, action/pose, environment, composition, lighting, style, texture, continuity, and constraints.
+- Target-model unknowns are routed to model-adaptation.
+- Negative handling matches model support.