@lemoncat7/dsh-partner 2.4.2 → 2.7.0

package/resources/skills/director-skills/upstream/skills/model-adaptation/tests/model_routing_tests.md

@@ -0,0 +1,8 @@
+# Model Routing Tests
+
+| Request | Expected behavior |
+|---|---|
+| Make this work in Kling | Load model-adaptation and Kling video profile |
+| Convert Grok prompt to Le Chat | Translate syntax, remove unsupported negatives |
+| Export for Sora and Hailuo | Use universal video brief, mark capabilities unknown |
+| Why did this work in Seedance but fail in Veo? | Compare duration, references, negatives, motion complexity |

package/resources/skills/director-skills/upstream/skills/model-adaptation/tests/prompt_translation_tests.md

@@ -0,0 +1,7 @@
+# Prompt Translation Tests
+
+- Source-specific parameters are not pasted blindly into target model.
+- Universal creative intent is preserved.
+- Negative prompt behavior is converted correctly.
+- Reference assets are mapped or marked unsupported/unknown.
+- Translation report lists changed, unsupported, and unknown fields.

package/resources/skills/director-skills/upstream/skills/model-adaptation/tests/research_provenance_tests.md

@@ -0,0 +1,6 @@
+# Research Provenance Tests
+
+- Public provenance documentation describes research categories without exposing private filenames.
+- Private local paths are not present in public docs.
+- Excluded unsafe or irrelevant material is noted at the category level.
+- New model research updates the relevant profile, confidence level, and unknown-capability tests.

package/resources/skills/director-skills/upstream/skills/model-adaptation/tests/unknown_capability_tests.md

@@ -0,0 +1,7 @@
+# Unknown Capability Tests
+
+When a profile lacks a capability:
+- Output must say `unknown`.
+- Do not infer from similar models.
+- Provide conservative fallback.
+- Recommend checking current official docs for live specs.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/SKILL.md

@@ -0,0 +1,82 @@
+---
+name: prompt-iteration-and-diagnostics
+description: "Diagnose failed AI image or video outputs, identify root causes, revise prompts with single-variable iteration, and produce structured prompt revision reports."
+---
+
+# prompt-iteration-and-diagnostics
+
+## When to use
+- The user shows or describes a failed output and asks how to fix it.
+- The user wants prompt diagnostics, iteration, A/B variants, failure analysis, or a revision report.
+- The output has drift, bad anatomy, wrong motion, weak realism, camera chaos, text/logo errors, prompt collapse, or model transfer failure.
+- The prose, dialogue, or film package has generic AI voice, weak story logic, rights/provenance gaps, or release-quality concerns.
+
+## When not to use
+- The user has no output or failure description and only needs first-draft prompting.
+- The issue is a tool outage or account/billing problem.
+- The user asks for unsafe bypass instructions.
+
+## Required inputs
+- Original prompt
+- Observed output or failure description
+- Target model/tool
+- Desired result
+
+## Optional inputs
+- Reference images
+- Settings/parameters
+- Seed/version
+- Previous attempts
+- Continuity bible
+
+## Workflow
+1. Restate the intended result and the actual failure.
+2. Classify failure: prompt ambiguity, contradiction, overload, model limitation, continuity drift, temporal overload, reference conflict, safety/policy rejection, or parameter mismatch.
+3. Identify the smallest change likely to improve the result.
+4. Apply the one-variable rule for iterative tests unless the prompt is fundamentally broken.
+5. Rewrite using the relevant skill: image, video, continuity, style, or model-adaptation.
+6. Produce a revision report with diagnosis, changed fields, unchanged anchors, expected improvement, and next test.
+7. When the failure is a model limitation, redesign the shot or route to a better model instead of forcing the same prompt.
+
+## Decision logic
+- If output is 80 percent correct, make minimal edits.
+- If identity drift appears, strengthen references and bible anchors.
+- If motion fails, reduce actions and camera moves.
+- If text/logos fail, choose a text-capable model or simplify typography.
+- If a prompt is rejected, remove unsafe/IP-sensitive content and do not provide bypass tactics.
+
+## Output formats
+- Failure diagnosis
+- Revised prompt
+- A/B test variants
+- Prompt revision report
+- Continuity update recommendations
+- Model-routing note
+- Story/voice/release QC report
+
+## Quality checks
+- Diagnosis maps to a specific prompt or model cause.
+- Revisions preserve what already worked.
+- Only meaningful variables change per test.
+- The report distinguishes fixable prompt issues from model limitations.
+- Story, voice, continuity, performance, rights, and provenance checks are separated when evaluating a full AI film package.
+- No unsafe bypass methods are included.
+
+## Anti-patterns
+- Rewriting the entire prompt after a small failure
+- Adding long negative lists without prioritization
+- Blaming the model before checking contradictions
+- Ignoring uploaded reference conflicts
+- Treating policy rejection as a prompt-engineering puzzle.
+
+## Exit criteria
+- The user has a revised prompt and a clear next test or model-routing decision.
+
+## Supporting files
+Read only the supporting file needed for the active task:
+- `references/iteration_diagnostics.md`
+- `references/failure_modes.md`
+- `references/revision_loops.md`
+- `references/story_voice_release_qc.md`
+- `templates/prompt_revision_report.md`
+- `templates/failed_output_intake.md`

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/checklists/diagnostic_checklist.md

@@ -0,0 +1,8 @@
+# Diagnostic Checklist
+
+- Intended vs actual result is stated.
+- Primary failure is classified.
+- Prompt issue vs model limitation is identified.
+- What worked is preserved.
+- One variable is changed for close outputs.
+- Unsafe bypasses are not provided.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/checklists/release_quality_checklist.md

@@ -0,0 +1,9 @@
+# Release Quality Checklist
+
+- Story: every scene advances conflict, character, or theme.
+- Voice: dialogue/prose avoids generic AI phrasing and on-the-nose emotion.
+- Continuity: character, costume, prop, location, light, and palette states match the bible.
+- Performance: dialogue is short enough to speak and sync naturally.
+- Post: color, grain, audio bed, subtitles, and upscaling are planned.
+- Rights: references, voices, likenesses, music, and stock assets are cleared for the intended use.
+- Provenance: prompts, settings, outputs, manual edits, and metadata are archived.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/examples/failed_output_to_revised_prompt.md

@@ -0,0 +1,10 @@
+# Example: Failed Output to Revised Prompt
+
+## Failure
+The video changed the key into a silver modern key and added extra fingers.
+
+## Diagnosis
+Reference preservation and hand anatomy failed. The prompt over-described mood but under-specified immutable prop and hand constraints.
+
+## Revision
+Use the uploaded image as the exact visual anchor. Preserve the worn brass key stamped 12B, same shape, same tarnished color, same guard hand with five fingers, same lock and chipped green door paint. Animate only a small tremble followed by one slow clockwise turn. Locked-off macro camera. Cool fluorescent light. Constraints: no key redesign, no extra fingers, no background morphing.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/examples/final_prompt_package_export.md

@@ -0,0 +1,10 @@
+# Example: Final Prompt Package Export
+
+## Package contents
+- Logline and treatment.
+- Scene list and shot list.
+- Character/location/prop bible.
+- Keyframe image prompts for S01-001 to S01-004.
+- I2V prompts for each shot.
+- Model-specific exports for Grok, Kling, and Seedance.
+- Iteration log and known risks: hands, low light, door text.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/references/failure_modes.md

@@ -0,0 +1,13 @@
+# Common Failure Modes
+
+| Symptom | Likely cause | First fix |
+|---|---|---|
+| Identity drift | no reference or weak anchors | use character sheet/reference workflow |
+| Waxy plastic look | missing texture/light physics | add skin/fabric/surface detail and physical shadows |
+| Camera chaos | too many moves | keep one dominant move |
+| Subject stops moving | action not continuous | specify continuous action for full duration |
+| Bad hands | close hand task | reframe wider, simplify interaction |
+| Text gibberish | model text limitation | choose text-capable model or simplify text |
+| Prompt ignored | important detail buried | front-load key instruction |
+| Over-stylized output | too many style anchors | choose one dominant style plus execution details |
+| Rejection | unsafe/IP/policy-sensitive content | remove or generalize sensitive content, do not bypass |

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/references/iteration_diagnostics.md

@@ -0,0 +1,17 @@
+# Iteration Diagnostics
+
+## Failure categories
+- Ambiguous subject or action.
+- Contradictory style, lighting, or camera directions.
+- Prompt overload.
+- Temporal overload for video duration.
+- Reference asset conflict.
+- Missing continuity anchors.
+- Model unsupported capability.
+- Safety, IP, or policy rejection.
+
+## One-variable rule
+If the output is mostly right, change one high-leverage variable at a time: subject wording, camera move, light source, reference image, negative constraint, aspect ratio, or duration.
+
+## When to abandon
+Start over or redesign when the failure is structural: impossible duration, central hand/mirror/text task in a weak model, too many characters, or unsupported feature dependency.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/references/revision_loops.md

@@ -0,0 +1,17 @@
+# Revision Loops
+
+## Minimal loop
+1. Define intended result.
+2. Observe actual result.
+3. Name the single failure that matters most.
+4. Change one variable.
+5. Predict the effect.
+6. Test and record.
+
+## Structured A/B loop
+- A: preserve prompt, adjust camera/lens.
+- B: preserve prompt, adjust light/palette.
+- C: preserve prompt, adjust reference/continuity.
+- D: preserve prompt, adjust model/settings.
+
+Use A/B only when the original output is close. For broken prompts, rebuild the universal brief first.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/references/story_voice_release_qc.md

@@ -0,0 +1,30 @@
+# Story, Voice, and Release QC
+
+Use this for full story/film packages, not just isolated prompts.
+
+## Story check
+
+- Does every scene advance conflict, character, theme, or necessary setup?
+- Does the ending pay off planted details rather than adding new explanation?
+- Is the scale appropriate for the target runtime?
+- Are beautiful AI-generated moments serving the story, or distracting from it?
+
+## Voice check
+
+- Does the prose/dialogue still sound specific, or has it become generic AI smoothness?
+- Are there cliches, overwritten feelings, or interchangeable speakers?
+- Do dialogue lines survive being read aloud?
+- Are sensory details specific to character and setting?
+
+## Continuity check
+
+- Are golden images, bible entries, shot prompts, and final clips using the same asset IDs?
+- Are prop, costume, time-of-day, and damage states consistent?
+- Did model transfer alter identity, palette, or setting geography?
+
+## Release hygiene
+
+- Document prompts, reference sources, model/tool versions, manual edits, and final selection decisions.
+- Confirm rights for stock, references, voices, likenesses, music, and generated outputs.
+- Preserve provenance metadata such as C2PA/SynthID where available.
+- Treat legal/copyright guidance as a risk checklist, not legal advice.

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/templates/failed_output_intake.md

@@ -0,0 +1,13 @@
+# Failed Output Intake Template
+
+- Target model/tool:
+- Original prompt:
+- Settings/parameters:
+- Reference assets:
+- Intended result:
+- Actual result:
+- What worked:
+- What failed:
+- Must preserve:
+- Can change:
+- Number of prior attempts:

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/templates/final_production_package.md

@@ -0,0 +1,11 @@
+# Final Production Package Template
+
+## Project brief
+## Story assets
+## Continuity bible
+## Shot list
+## Image prompts
+## Video prompts
+## Model-specific exports
+## Iteration notes
+## Known risks and fallback plans

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/templates/prompt_revision_report.md

@@ -0,0 +1,17 @@
+# Prompt Revision Report Template
+
+## Diagnosis
+- Primary failure:
+- Likely cause:
+- Model limitation or prompt issue:
+
+## Preserved elements
+
+## Changed elements
+
+## Revised prompt
+
+## Test plan
+- Variable changed:
+- Expected improvement:
+- Next fallback if it fails:

package/resources/skills/director-skills/upstream/skills/prompt-iteration-and-diagnostics/tests/prompt_quality_checklist_tests.md

@@ -0,0 +1,6 @@
+# Prompt Quality Checklist Tests
+
+- Failure diagnosis distinguishes prompt issue from model limitation.
+- Revised prompt preserves working elements.
+- One-variable iteration is used for close outputs.
+- Unsafe bypass instructions are excluded.

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/SKILL.md

@@ -0,0 +1,81 @@
+---
+name: screenplay-and-scene-writing
+description: "Write and revise screenplay scenes, dialogue, beats, emotional turns, pacing, visual action, and screenplay-style excerpts from short-film concepts or outlines."
+---
+
+# screenplay-and-scene-writing
+
+## When to use
+- The user asks for screenplay pages, scene writing, dialogue, beats, pacing, emotional turns, or visual storytelling.
+- The user provides a treatment, outline, scene purpose, or character conflict and wants it dramatized.
+- The task is to revise a scene for clarity, subtext, tension, or cinematic action.
+
+## When not to use
+- The user only needs shot list extraction from already final pages.
+- The user asks for image/video model parameters.
+- The user wants a full prose-fiction workflow rather than screenplay or scene design; route early concept work to `short-film-development`.
+
+## Required inputs
+- Scene purpose
+- Characters present
+- Location and time
+- Before/after story state
+- Desired tone or genre
+
+## Optional inputs
+- Dialogue constraints
+- Runtime/page target
+- Visual bible anchors
+- Specific format requirements, including Fountain
+
+## Workflow
+1. Define the scene transaction: who wants what, who resists, what changes.
+2. List beats as playable actions, not explanations.
+3. Choose the dominant point of view and what the audience knows at the start.
+4. Write action lines as visible and audible behavior with concise sensory detail.
+5. Write dialogue with subtext: characters pursue tactics, avoid saying the theme directly, and reveal pressure through choices.
+6. If the user needs importable screenplay text, output strict Fountain markup and keep camera/production notes outside the scene.
+7. Add an emotional turn and a final image or button that changes momentum.
+8. Revise for compression: remove repeated beats, expository greetings, and internal states that cannot be filmed.
+
+## Decision logic
+- If dialogue carries exposition, convert part of it to behavior, prop use, interruption, or silence.
+- If a beat cannot be shot, rewrite it as a visible action or sound.
+- If the scene has no reversal, add discovery, choice, escalation, or cost.
+- If prompts are a downstream goal, tag key visual anchors after the scene.
+
+## Output formats
+- Beat sheet
+- Screenplay scene excerpt
+- Fountain scene
+- Dialogue pass
+- Visual action pass
+- Scene revision notes
+- Prompting handoff notes
+
+## Quality checks
+- Every line changes tactic, pressure, or information.
+- Action lines are filmable and not novelistic.
+- Characters sound distinct by goal, rhythm, and avoidance strategy.
+- The scene has a clear entry state, turn, and exit state.
+- Important continuity details are captured for the bible.
+
+## Anti-patterns
+- Explaining backstory before the conflict starts
+- Writing camera directions inside a draft unless requested
+- Dialogue that states exactly what characters feel
+- Unfilmable internal adjectives without behavior
+- Scenes that end where they began
+
+## Exit criteria
+- A scene or beat sheet is ready for shot breakdown, revision, or prompt generation.
+
+## Supporting files
+Read only the supporting file needed for the active task:
+- `references/scene_construction.md`
+- `references/dialogue_and_subtext.md`
+- `references/screenplay_format.md`
+- `references/fountain_workflow.md`
+- `templates/screenplay_scene.md`
+- `templates/fountain_scene.md`
+- `templates/beat_sheet.md`

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/checklists/scene_quality_checklist.md

@@ -0,0 +1,8 @@
+# Scene Quality Checklist
+
+- Scene starts under pressure.
+- Every beat changes tactic, knowledge, or emotion.
+- Action lines are filmable.
+- Dialogue has subtext.
+- Exit state differs from entry state.
+- Continuity changes are recorded.

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/examples/scene_to_screenplay_excerpt.md

@@ -0,0 +1,18 @@
+# Example: Scene to Screenplay Excerpt
+
+```text
+INT. BASEMENT HALLWAY - NIGHT
+
+The guard's flashlight trembles over a row of storage doors. The song leaks from 12B, thin and warped.
+
+He fits a brass key into the lock. Stops.
+
+GUARD
+You picked a terrible night.
+
+No answer. The cassette warbles, slower now, as if the batteries are dying.
+
+He turns the key.
+```
+
+Handoff: storage door 12B, brass key, warped cassette audio, flashlight tremor, blue-green fluorescent light.

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/references/dialogue_and_subtext.md

@@ -0,0 +1,15 @@
+# Dialogue and Subtext
+
+## Dialogue does
+- Pursues a tactic.
+- Conceals or reveals pressure.
+- Changes relationship distance.
+- Buys time, attacks, seduces, deflects, tests, pleads, or traps.
+
+## Dialogue avoids
+- Explaining theme.
+- Saying exactly what the character feels unless collapse or confession is the beat.
+- Repeating information already visible.
+
+## Revision pass
+For each line, label the tactic. If two lines perform the same tactic, cut or combine one.

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/references/fountain_workflow.md

@@ -0,0 +1,29 @@
+# Fountain Workflow
+
+Use Fountain when the user wants plain-text screenplay output that can be imported into screenwriting tools.
+
+## Core syntax
+
+```text
+INT. LOCATION - NIGHT
+
+Action lines describe visible and audible behavior in present tense.
+
+CHARACTER
+Dialogue appears under the character name.
+
+(brief parenthetical only when needed)
+Dialogue continues.
+```
+
+## Rules for AI-assisted drafting
+
+- Write one scene or beat at a time to preserve continuity and voice.
+- Keep action lines filmable; do not include internal thoughts unless represented by action, sound, or image.
+- Use parentheticals sparingly. They guide performance; they are not prose commentary.
+- Keep camera directions out of the screenplay unless the user explicitly requests a shooting script.
+- Put visual anchors, prompt notes, and shot ideas in a separate handoff block after the Fountain scene.
+
+## Page/runtime note
+
+Standard screenplay pages roughly map to screen minutes, but AI shorts should still be planned by beats and shot duration because generated clips have model-specific time limits.

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/references/scene_construction.md

@@ -0,0 +1,20 @@
+# Scene Construction
+
+## Scene transaction
+Every scene should answer:
+- Who wants what right now?
+- Who or what resists?
+- What tactic is used first?
+- What changes the tactic?
+- What is gained, lost, revealed, or damaged?
+
+## Beat workflow
+1. Entry state.
+2. First tactic.
+3. Resistance/interruption.
+4. Escalation or reveal.
+5. Choice under pressure.
+6. Exit image or changed state.
+
+## Visual writing
+Write what can be seen or heard. Replace "she feels guilty" with behavior: she wipes the clean glass again, avoids the photo, or pockets the receipt.

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/references/screenplay_format.md

@@ -0,0 +1,18 @@
+# Screenplay Format
+
+Use standard readable screenplay conventions when requested:
+
+```text
+INT. LOCATION - TIME
+
+Action lines in present tense, concise and filmable.
+
+CHARACTER
+Dialogue line with subtext and pressure.
+```
+
+For AI production, include a post-scene handoff block only outside the screenplay excerpt:
+- Visual anchors
+- Continuity changes
+- Candidate shots
+- Image/video prompt needs

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/templates/beat_sheet.md

@@ -0,0 +1,6 @@
+# Scene Beat Sheet Template
+
+| Beat | Character tactic | Resistance | Subtext | Visible action | Turn |
+|---|---|---|---|---|---|
+| 1 | | | | | |
+| 2 | | | | | |

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/templates/dialogue_pass.md

@@ -0,0 +1,9 @@
+# Dialogue Pass Template
+
+For each line:
+- Speaker:
+- Literal line:
+- Tactic:
+- What they avoid saying:
+- Pressure underneath:
+- Cut/compress/rewrite note:

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/templates/fountain_scene.md

@@ -0,0 +1,25 @@
+# Fountain Scene Template
+
+```text
+INT./EXT. LOCATION - TIME
+
+[Visible present-tense action. Introduce the pressure quickly.]
+
+CHARACTER
+[Playable dialogue with subtext.]
+
+[Action or interruption that changes the beat.]
+
+CHARACTER
+[Short response, tactic, evasion, or reveal.]
+
+[Final image, decision, or new problem.]
+```
+
+## Handoff block
+- Scene purpose:
+- Entry state:
+- Exit state:
+- Visual motif:
+- Continuity changes:
+- Candidate shot-list beats:

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/templates/screenplay_scene.md

@@ -0,0 +1,19 @@
+# Screenplay Scene Template
+
+```text
+INT./EXT. LOCATION - TIME
+
+[Present-tense visual action. Keep lines short and filmable.]
+
+CHARACTER
+[Dialogue with tactic and subtext.]
+
+[Action/reaction that changes the beat.]
+```
+
+## Post-scene handoff
+- Entry state:
+- Exit state:
+- Visual anchors:
+- Continuity changes:
+- Candidate shots:

package/resources/skills/director-skills/upstream/skills/screenplay-and-scene-writing/tests/screenplay_scene_tests.md

@@ -0,0 +1,6 @@
+# Screenplay Scene Tests
+
+- Given a scene purpose, output uses screenplay-style slug/action/dialogue when requested.
+- Action lines are visible/audible.
+- Dialogue lines have tactics or subtext.
+- Post-scene handoff is outside the screenplay excerpt.

package/resources/skills/director-skills/upstream/skills/short-film-development/SKILL.md

@@ -0,0 +1,83 @@
+---
+name: short-film-development
+description: "Develop short stories and short films from raw ideas into loglines, premises, themes, character engines, conflict, structure, treatments, scene lists, prose plans, and revision plans for cinematic AI production."
+---
+
+# short-film-development
+
+## When to use
+- A user has a raw idea, image, premise, theme, genre, character, or mood and wants a short-film concept.
+- The user asks for loglines, treatments, outlines, short-story plans, character conflict, theme, acts, or a revision plan.
+- The next downstream task is screenplay, shot list, visual bible, image prompts, or video prompts.
+
+## When not to use
+- The user already has locked script pages and only needs shot extraction.
+- The task is only model-specific prompt translation.
+- The user wants legal clearance, financing, casting, or production scheduling.
+
+## Required inputs
+- Raw idea or constraint
+- Target length or runtime range
+- Genre/tone/audience if known
+- Any required characters, locations, or production limits
+
+## Optional inputs
+- Reference films or style anchors
+- Theme or emotional promise
+- Available AI models or output format
+- Continuity bible entries
+
+## Workflow
+1. Clarify the creative intent layer: story engine, theme, protagonist desire, emotional wound, obstacle, stakes, point of view if prose, and audience experience.
+2. Generate 3 to 5 divergent concept routes when the idea is early; select or recommend one based on specificity, feasibility, and cinematic promise.
+3. Write a logline with protagonist, goal, obstacle, stakes, and irony or pressure.
+4. Expand into premise, theme statement, character engine, central conflict, world rules, and ending direction.
+5. Choose a compact structure: unity-of-effect short story, Fichtean crisis curve, single turning point, three-beat micro-arc, circular reveal, escalation spiral, or procedural countdown.
+6. Create a scene list where every scene has dramatic purpose, visual purpose, emotional turn, and continuity needs.
+7. Produce a revision plan that strengthens causality, compresses exposition, and removes unfilmable abstraction.
+
+## Decision logic
+- If the idea is vague, run idea intake before writing.
+- If the concept is visually strong but narratively thin, build the character want/need conflict first.
+- If runtime is under 3 minutes, prefer one location, one core conflict, and one major turn.
+- If the user wants prompts soon, preserve image/video-ready visual anchors in every scene.
+
+## Output formats
+- Logline
+- Premise sheet
+- Theme and audience experience brief
+- Character conflict map
+- Short-story plan
+- Short treatment
+- Scene list
+- Revision plan
+- Downstream handoff brief
+
+## Quality checks
+- The logline contains a filmable conflict, not just a mood.
+- Each scene changes story state or emotional state.
+- The ending answers the opening tension.
+- The concept includes reusable visual anchors for continuity and prompting.
+- The treatment separates what the audience sees/hears from backstory.
+
+## Anti-patterns
+- Abstract themes with no visible action
+- Too many locations for a short runtime
+- Worldbuilding that does not affect choices
+- A protagonist who only observes
+- Prompt-ready imagery that contradicts the story tone
+
+## Exit criteria
+- A concise concept package exists and can feed screenplay, shot list, bible, or prompt generation without re-interviewing the user.
+
+## Supporting files
+Read only the supporting file needed for the active task:
+- `references/story_development_workflow.md`
+- `references/ai_storytelling_workflow.md`
+- `references/screenplay_structure.md`
+- `references/continuity_handoff.md`
+- `templates/idea_intake.md`
+- `templates/logline.md`
+- `templates/short_film_treatment.md`
+- `templates/short_story_plan.md`
+- `templates/story_revision_audit.md`