Communitygithub.com

agbelemi/google-flow-scripting-skill

Google Flow Prompt Guide: Veo Character Consistency, Storyboarding and Continuity

google-flow-scripting-skill 是什么?

google-flow-scripting-skill is a Claude Code agent skill that google Flow Prompt Guide: Veo Character Consistency, Storyboarding and Continuity.

兼容平台~Claude Code~Codex CLI~Cursor
npx skills add agbelemi/google-flow-scripting-skill

在你喜欢的 AI 中提问

打开一个已预加载此 Agent Skill 的新对话。

文档

Google Flow Scripting & Prompting

Version: 1.3.1

Use this Skill to turn a video brief into a production-ready, continuity-aware package for Google Flow or a comparable video generator.

Operating principles

  1. Do not assume implicit memory between independent generations. Carry continuity explicitly through self-contained prompt details, reference images or ingredients, saved frames, first/last-frame controls, extensions, and state tracking.
  2. Treat platform limits as dated facts, not eternal rules. Read reference/FLOW-FEATURES.md before making a capability claim. When the current interface disagrees, trust the interface and flag the discrepancy.
  3. Plan assets before final segment prompts. Establish the cast, recurring objects, environments, reference priority and set anchors before writing detailed shots.
  4. Storyboard before expensive video generation. Check identity, geography, composition, state and handoffs in stills first.
  5. Make each generation prompt usable on its own. Avoid phrases such as “the same person,” “as before,” or “continuing from scene 2” unless an actual attached reference or frame is named explicitly.
  6. Use project policies, not universal creative laws. One beat per second, continuous shots, no on-screen text and generated audio are defaults that can be changed when the brief requires something else.

Workflow

1. Establish the brief

Extract what is already known before asking questions. Resolve only production-critical gaps:

  • format and visual treatment
  • target runtime and segment duration
  • aspect ratio and delivery platform
  • audience and objective
  • dialogue, generated audio, intentional silence, or sound added in post
  • on-screen text policy
  • rights-sensitive people, brands, locations or claims

Do not repeat questions the user has already answered. Make visible assumptions for minor gaps.

2. Choose the relevant craft files

Read:

  • core/flow-story-architect.md for story structure and state tracking
  • core/flow-asset-manager.md for references, ingredients and set anchors
  • core/flow-storyboard-director.md for the storyboard package
  • the closest file in formats/ for format-specific craft
  • core/flow-continuity-auditor.md before approving the final script

Use core/flow-orchestrator.md as the end-to-end routing reference.

3. Build the production package

Produce, in this order:

  1. Brief and explicit assumptions
  2. Runtime arithmetic and segment map
  3. Cast, object, environment and state inventory
  4. Prioritised reference/ingredient plan
  5. Storyboard or keyframe plan
  6. Self-contained segment prompts
  7. Handoff map showing cut, match cut, continuation, extension, or first/last-frame strategy
  8. Mechanical validation results
  9. Human continuity audit and unresolved risks

4. Prompt text policy

The first semantic line of each prompt must declare one of these:

NO TEXT IN THE IMAGE: do not render words, labels, captions or watermarks.

or:

INTENTIONAL TEXT IN THE IMAGE: state the exact words, surface and location; no other text.

Do not use a blanket no-text rule when packaging, signs, interfaces, subtitles or titles are intentionally part of the shot.

5. Audio policy

For generated-audio workflows, include an AUDIO: block with dialogue, one primary sound effect, ambience and exclusions when useful.

Deliberate silence is explicit:

AUDIO: Intentional silence.

For visual-only previs or sound added entirely in post, say so in the project assumptions and validate with --no-require-audio.

6. Validate

Run:

python scripts/validate.py your-script.md --segment-length 8

For visual-only or post-audio workflows:

python scripts/validate.py your-script.md --segment-length 8 --no-require-audio

The validator is a mechanical linter, not a substitute for judgement. After it passes, apply the continuity auditor to reference coverage, spatial continuity, state continuity, handoffs and final-frame feasibility.

Safety and rights

Before using a real person, protected brand asset, sensitive claim, archival material or documentary reconstruction, identify the consent, licence, disclosure and factual-verification requirements. Never present generated reconstruction as authentic evidence.

Supporting resources

  • reference/PLAYBOOK.md — full production methodology
  • reference/FLOW-FEATURES.md — dated platform capability notes
  • reference/TEMPLATES.md — prompt and deliverable templates
  • reference/FAILURE-MODES.md — evidence-tagged observed and inferred failures
  • reference/ADAPTING-OTHER-GENERATORS.md — portability notes
  • reference/VERIFICATION.md — what “documentation reviewed” means for this release

相关技能