Skip to content

agentsope/SkillAlchemy / skills/agentsope-ai-video-director

agentsope-ai-video-director

Unverified

Direct and implement programmatic videos from a natural-language sentence: brief, shots, style, timeline, assets, engine-specific coding prompts, draft, review and repair. Use when asked to make a video, product launch, explainer, Reel/Short, kinetic typography or docs-to-video with HyperFrames, Remotion or Motion Canvas, or to improve an existing code-video project. 一句话视频制作与导演工作流。

# agentsope-ai-video-director

从一句话生成可执行导演方案、独立引擎编码提示,并在工具可用时实施、渲染、检查和修订视频。默认输出真实达到的状态;方案、代码、编码文件、看过完整动态和听过声音是不同结果。

Use for product/SaaS launches, vertical social clips, data/technology explainers, kinetic typography/brand, cinematic visual shorts, website/GitHub/docs videos and voice-led knowledge videos. These seven application guides are authored package structures. Their research support varies; cinematic coverage is limited and diffusion procedures have not been distilled.

Start with [SKILL.md](SKILL.md). Example: “用 15 秒展示一个虚构任务管理产品的三步整理流程,用原创图形,先做可渲染草稿。” Critical product claims, assets and rights stay pending; reversible aesthetic defaults are declared assumptions. The agent continues authorized provisional work and creates actual code/video where an eligible exporter exists.

## Project-local installation

Place this runtime directory at `<project>/dist/agentsope-ai-video-director/` (or another project-local location). Link only within that project:

```sh
mkdir -p .agents/skills .claude/skills
ln -s ../../dist/agentsope-ai-video-director .agents/skills/agentsope-ai-video-director
ln -s ../../dist/agentsope-ai-video-director .claude/skills/agentsope-ai-video-director
```

If discovery links already exist, inspect their target before replacing them. This guide does not install engines or global skills. Engine dependencies belong to the video project, and must be probed/version-matched before rendering.

## Engines and verification limits

HyperFrames 0.8.142 uses HTML composition seconds and completed paused timeline registration. Remotion 4.0.534 uses deterministic frame state, with separate Node/server and browser-compositor adapters. Motion Canvas released 3.17.2 uses seconds in generators and an editor-driven registered exporter; no headless CLI is claimed. Canvas/WebGL remains a source-local documented renderer note, not a universal fallback. Read [engine adapters](references/engine-adapters.md).

Research-time original HyperFrames and Remotion silent 2-second smoke videos were encoded, probed and sampled. They do not validate voice, all media/fonts, full pacing or listening. Motion Canvas TypeScript/Vite/server checks passed; actual video export/UI/audio remain unexecuted. Browser Remotion export is also unexecuted. Remotion production eligibility remains user/use-specific under pinned v4 terms. The inspected cached Motion Canvas exporter FFmpeg binary had nonfree build terms and is excluded from distribution. Exact release/main revision and binary terms must be checked separately.

## Examples and checks

All examples are original synthetic demonstrations, never counted research cases; their represented video checks remain unexecuted until an actor performs them.

- [Sentence to plan](examples/sentence-to-director-plan.md)
- [Detailed HyperFrames prompt](examples/hyperframes-prompt.md)
- [Detailed Remotion prompt](examples/remotion-prompt.md)
- [15s product application](examples/product-launch.md)
- [30s data application](examples/data-explainer.md)
- [60s narrated docs application](examples/narrated-docs.md)
- [Sentence fixtures](tests/sentence-inputs.json), [routing fixtures](tests/engine-routing.json), [missing/error fixtures](tests/missing-and-errors.json)
- [Quality checklist](tests/quality-checklist.md), [manual review and output contract](tests/manual-review.md)
- [Package validator](tests/validate_package.py), [actual-output validator](tests/validate_outputs.py)

Run from any working directory using the actual paths:

```sh
python3 <skill-dir>/tests/validate_package.py
python3 <skill-dir>/tests/validate_outputs.py --outputs <actual-agent-outputs.json>
```

These checks establish structure and recorded contract compliance, not full audiovisual quality. Validators do not produce canned agent answers. Retain real agent outputs, commands and observed checks separately; label unexecuted tests honestly.

## Reference index

- [Prompt patterns](references/prompt-patterns.md), [director playbook](references/director-playbook.md), [seven video types](references/video-types.md)
- [Quality rubric](references/quality-rubric.md), [40 operation cards](references/sop_models.md)
- [Sources and rights](references/sources.md), [42 creative case ledger](references/case-studies.md)
- [Research coverage](references/research_notes.md), [rule/finding/source provenance](references/provenance.json)
- [Metadata and explicit runtime manifest](skill.json), [original-content MIT license](LICENSE)

The runtime manifest has 27 authored files. Local research reports, caches, tooling, Git history, installed dependencies/binaries and exemplar copies are audit material excluded from installation/archive. Every important operation traces to an admitted P rule, exact scope, finding and source/version. C01–C09 are separately declared authored contracts. General admission does not mean universal creative effectiveness; P02/P05/P06/P09 are optional author suggestions awaiting empirical validation.

## Executed package evaluation — additive release record

A fresh delegated worker loaded this compiled skill and produced 14 retained actual responses: the three sentence requests (P01/P02/P03), five routing cases and six missing/error cases. The strengthened actual-output validator passed the final retained response set (exit 0). The first failure, exact-fixture clarification and follow-up reruns remain in the separate local audit. These are delegated instruction-workflow evaluations, not external Claude Code or Codex CLI integration tests. The authored fixture definitions retain their original `not_executed` labels; the executed responses and reports are recorded independently.

The root also implemented and checked an independent original B01 HyperFrames benchmark: four shots, 15 seconds, 1080×1920, 30 fps and 450 decoded frames; H.264/yuv420p, silent MP4, 1,410,286 bytes. It differs from the worker’s five-shot 720×1280 P01 response and is excluded from research-case counts. Its actual technical probe passed. Browser checks reported zero runtime/layout errors and 42/42 contrast checks, with five structural warnings and motion audit disabled. Observed implementation, font and shot-boundary defects were fixed and rechecked using before/after evidence. The root inspected eight representative frames and twelve boundary frames; four ordered/shuffled seek samples had identical SHA-256 hashes.

Full-motion viewing remains **not_executed**. Audio listening is **not_applicable only for this explicitly silent benchmark**; it does not clear pending narration in P03 or the narrated-docs application. Final production rights and product claims remain unknown, and `publish_ready` is false. The encoded draft, frames, logs and reviews remain in the separate local audit, outside the runtime archive. Static/technical/frame evidence establishes only its actual coverage and does not certify audiovisual or production completion.

Two independent reviews completed: evidence/rights/scope passed with limitations, and activation/routing/runtime review passed with nonblocking limits. The route review strengthened all authored route/error contracts and the R08 separate-engine Prompt bundle. Twenty-one checker mutation probes were executed and rejected their deliberately invalid contracts; they are validator checks, not actor responses. The historical research observations above remain unchanged.
rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).Stars: 198kUpdated Claude CodeCodexUnverifiedInstall
doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.Stars: 198kUpdated Claude CodeCodexVerifiedInstall