원클릭으로 Manus에서 모든 스킬 실행

시작하기

dircreative-longform-reference-planner

스타2

포크0

업데이트2026년 7월 14일 14:49

Build global and per-sequence reference pack contracts for longform AI video workflows.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

papperrollinggery

papperrollinggery/Paperrolling-DIRcreative-SKILL

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

Longform Reference Planner

Required Knowledge

docs/film-preproduction/longform-decomposition-policy.md
docs/film-preproduction/capability-aware-generation-policy.md
docs/film-preproduction/reference-locking-policy.md
docs/film-preproduction/schemas/longform-reference-pack.yaml
docs/film-preproduction/schemas/reference-pack-manifest.yaml
docs/film-preproduction/research/model-reference-behavior.md

Inputs

sequence plan
visual bible
reference pack manifest
image prompt manifest
target model list

Outputs

longform reference pack
global reference pack
per-sequence reference pack plan
clean start/end frame requirements
external generation instructions

Rules

Separate global identity/style assets from sequence assets.
Use sequence packs as the normal unit for 180s work.
Do not make one clean image per shot unless QA or model constraints require it.
Every asset must have one primary job and one asset_output.status.
prompt_only must remain fully useful without generated files.
assisted_generation requires recorded user authorization.
external_generation must include upload/import instructions.
Resolve every target through capability_card_id + version + provider_surface; never emit family aliases or family-keyed policy maps.
Copy no numeric reference limit into the pack; validate limits from the exact current card at audit time.
Keep dense boards out of direct frame slots whenever the exact card requires a clean image input.
Use clean first/end frames only for exact cards whose reference modes authorize those inputs.
Do not generate images or videos.

skill_run_receipt

Record global packs, sequence packs, output statuses, user gates, exact-card direct input policy, QA status, and next_recommended_skill: image-prompt-compiler.

이 저장소의 다른 Skills

같은 저장소

dircreative

papperrollinggery/Paperrolling-DIRcreative-SKILL

Use only after the user explicitly invokes $dircreative for film preproduction: story, script, storyboard, shot planning, visual systems, reference-image planning, and model-specific video prompts. ADCO mode accepts only a validated Specialist Exchange handoff. Do not use for maintaining, debugging, refactoring, testing, or evaluating the DIRcreative source repository; maintaining ADCO; ordinary code work; factual questions; or generic advertising requests without an explicit $dircreative invocation.

2026-07-212

dircreative-chat-facilitator

papperrollinggery/Paperrolling-DIRcreative-SKILL

Present DIRcreative work result-first and stop only at a real v2 external decision.

2026-07-192

dircreative-director-room

papperrollinggery/Paperrolling-DIRcreative-SKILL

Select dynamic narrative, visual-production, and model-continuity perspectives for Studio film-preproduction work.

2026-07-192

dircreative-story-development

papperrollinggery/Paperrolling-DIRcreative-SKILL

Convert selected concept into logline, beat sheet, treatment, and story review.

2026-07-192

dircreative-visual-bible

papperrollinggery/Paperrolling-DIRcreative-SKILL

Define visual continuity for characters, environments, props, wardrobe, lighting, palette, and texture.

2026-07-192

dircreative-checkpoint

papperrollinggery/Paperrolling-DIRcreative-SKILL

Save and resume workflow state across phases, artifacts, and skill runs.

2026-07-142

name	dircreative-longform-reference-planner
description	Build global and per-sequence reference pack contracts for longform AI video workflows.

Longform Reference Planner

Required Knowledge

docs/film-preproduction/longform-decomposition-policy.md
docs/film-preproduction/capability-aware-generation-policy.md
docs/film-preproduction/reference-locking-policy.md
docs/film-preproduction/schemas/longform-reference-pack.yaml
docs/film-preproduction/schemas/reference-pack-manifest.yaml
docs/film-preproduction/research/model-reference-behavior.md

Inputs

sequence plan
visual bible
reference pack manifest
image prompt manifest
target model list

Outputs

longform reference pack
global reference pack
per-sequence reference pack plan
clean start/end frame requirements
external generation instructions

Rules

Separate global identity/style assets from sequence assets.
Use sequence packs as the normal unit for 180s work.
Do not make one clean image per shot unless QA or model constraints require it.
Every asset must have one primary job and one asset_output.status.
prompt_only must remain fully useful without generated files.
assisted_generation requires recorded user authorization.
external_generation must include upload/import instructions.
Resolve every target through capability_card_id + version + provider_surface; never emit family aliases or family-keyed policy maps.
Copy no numeric reference limit into the pack; validate limits from the exact current card at audit time.
Keep dense boards out of direct frame slots whenever the exact card requires a clean image input.
Use clean first/end frames only for exact cards whose reference modes authorize those inputs.
Do not generate images or videos.

skill_run_receipt

Record global packs, sequence packs, output statuses, user gates, exact-card direct input policy, QA status, and next_recommended_skill: image-prompt-compiler.