Multi-Model Shot Prompts

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites: none for the install itself: no account, no API key, no paid service, and nothing else to install first. For using the skill afterwards, read this first: the skill is text only — a single SKILL.md carrying a prompt skeleton, a camera-movement vocabulary, adaptation rules and a batch-rewrite template; it has no script, installs nothing, and needs no local dependency at all (no Python, Pillow, NumPy or ffmpeg). It only writes video-generation prompts. Actually generating a video is a separate step, outside this skill, on an external video model I choose myself (the models named in SKILL.md are Seedance, Kling (可灵), Veo, Runway, Hailuo (海螺), Wan (通义万相), Sora and Pika, or any other video model): those services are not part of this install, the skill gives no price, duration limit, resolution or frame rate for any of them — SKILL.md says expressly not to invent those and to check each model's official documentation / console — and any account or payment they require is mine to arrange with that provider. SKILL.md also mentions, as an option only, that many video models can be called through an OpenAI-compatible or aggregation gateway configured as BASE_URL: this skill has no script that sets or uses any gateway, and nothing in this install configures one. Install "multi-model-shot-prompts" for me. It rewrites one shot list / storyboard into video-generation prompts tailored to different video generation models, model-neutral and without invented model prices or parameters: apply a universal skeleton (subject + PRODUCT LOCK, subject action, environment/scene, camera movement, shot size/angle, light/tone, style, pace/duration description, plus constraints: keep the product consistent, generate no text, show no other brand, natural hands), using writing principles (concrete verbs; one action and one camera movement per shot; camera terms in a fixed position in general wording; the product described by noun + colour + material; for image-to-video, write only what changes), a nine-term camera vocabulary (slow push-in / dolly in, slow pull-out / dolly out, truck left / right, orbit 30° around the product, crane down / top-down descending, tracking shot following the hand, locked-off static camera, macro close-up with shallow depth of field, quick punch-in), and adaptation rules for the input mode of each shot (text-to-video, first-frame image-to-video, or first-and-last-frames). Per shot it outputs a Chinese version, an English version, a Negative line if the model supports negative prompts, and a local Fallback motion (for example motion "zoom_in" in storyboard.json); it also gives a template for batch-rewriting a whole shot list with a large language model. Repository: https://github.com/xianyu110/ecommerce-video-skills (the skill is the whole skills/multi-model-shot-prompts/ folder at the pinned commit: SKILL.md only, 1 file) Pinned version: install the code at commit 96c49b349b7d975bda8f05ce6669c1d33500c269 exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/xianyu110/ecommerce-video-skills into a temporary folder, check out commit 96c49b349b7d975bda8f05ce6669c1d33500c269, and copy the skills/multi-model-shot-prompts/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-19: [FINGERPRINTS] - SKILL.md: SHA-256 ea21d23dcf73ce8f3d878210a2f0a0fe52fb3e667735b9101b9416ad19d70b6c If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full (it is the only file in the folder; there is no script to read), and tell me in plain words: what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The skill folder holds 1 file only: SKILL.md (5,598 bytes). There is no scripts/ subfolder and no code at all in the folder. The skill is self-contained in its own folder and needs no file from elsewhere in the repository — not the other 12 skills, not the root scripts/, not the README, and not the repository requirements.txt at install time. Two cross-references inside SKILL.md are content references, not file dependencies: the shot list (storyboard) may come from the selling-point-storyboard skill, and the PRODUCT LOCK (the product-consistency description) is described in the image-to-video-shots skill — I can supply my own shot list and my own PRODUCT LOCK description without installing either of those skills. - Nothing is sent anywhere during installation itself; the only download at install time is the skill folder from GitHub at the pinned commit. No script is run during installation, no key is asked for, and there is no telemetry in the folder. - What leaves this machine at use time, disclosed here before it ever happens: the skill itself sends nothing to anyone. It has no code, makes no network call, reads no key, and defines no API call of its own. Using it is a separate step that only starts when I ask for shot prompts, and what it produces is prompt text, between you and me. A send happens only if and when I take that finished prompt text — and, depending on the mode, images — to an external video generation model I choose myself: the models named in SKILL.md are Seedance, Kling (可灵), Veo, Runway, Hailuo (海螺), Wan (通义万相), Sora and Pika, or any other video model. What I would paste or upload to that chosen provider is exactly this: the rewritten prompt text itself (subject + PRODUCT LOCK — that is, a description of my product's shape, colour, material and logo — plus the action, the environment/scene, the camera movement, the shot size/angle, the light/tone, the style and the constraints), and, by mode: for text-to-video, that text alone; for first-frame image-to-video, my first-frame image as well; for first-and-last-frames, both of those images; and, if the model supports a reference image / subject reference, my white-background main product photo, uploaded as the reference, with the prompt saying "the product from the reference". That paste / upload is a send to the provider I choose, under that provider's own terms — the skill does not do it, does not automate it, and never contacts any of those models itself. The batch-rewrite template in SKILL.md works the same way: it is text I fill in (target model, input mode, PRODUCT LOCK, and my shot list pasted where it says to paste it) and use with you in our conversation. - One optional mention in SKILL.md is not a function of this skill: its "calling method (optional)" note says many video models can be called through an OpenAI-compatible or aggregation gateway, that this repository's scripts bind to no vendor, and that a gateway address could be configured as BASE_URL (it points to an example in the README under a heading that does not appear under that exact title at the pinned commit; the README, which is outside the installed folder, gives under "Related" a tryallapi.com example, marked as a referral link and as unaffiliated, with OPENAI_BASE_URL=https://tryallapi.com/v1, and states that nothing in the repository requires it). This skill contains no script, sets no BASE_URL, and sends nothing to any gateway; do not configure any gateway as part of this install or of using this skill's prompt-writing. - The local-only parts at use time: applying the universal skeleton, the camera vocabulary, the adaptation rules (full skeleton + detailed PRODUCT LOCK for text-to-video; for first-frame image-to-video, drop scene/appearance description and keep only action + camera movement + constraints; for first-and-last-frames, describe only what happens between the two frames; negative wording — deformed product, changed logo, warped text, extra fingers, flicker, watermark — only if the model supports negative prompts, otherwise the same constraints written as positive sentences such as "the product keeps exactly the same shape, colour and logo"), the output format, the batch template and the checklist in SKILL.md — all of it is text work between you and me, with no code execution. Rules from SKILL.md: one action and one camera movement per shot, every shot carries a PRODUCT LOCK or "write only what changes", Chinese and English versions with the constraints also in positive form, a local fallback motion for every shot, fixes for a model that "redraws" the product, an ineffective camera move or an overdone action — and never any unverified model price, duration limit, resolution or other parameter. - Do not rewrite any shot list, do not paste or upload anything to any video model, do not configure any BASE_URL or gateway, and do not create any external account during this install: using the skill is my decision, later, when I actually ask for shot prompts. 3. When I say "yes, install", install it as a skill: copy the whole multi-model-shot-prompts folder into the agent's skills directory, in a folder named "multi-model-shot-prompts", keeping its 1 file (SKILL.md at its root). A manual clone followed by copying the skills/multi-model-shot-prompts folder into ~/.claude/skills/ is the route described here; any other route is fine only as long as every installed file matches its fingerprint. 4. Verify with no network calls: frontmatter valid (name: multi-model-shot-prompts), the 1 file in place with a matching fingerprint. There is no script to run for verification — do not run anything. 5. Report what was installed, where, and what I still need to do myself (for example: provide a shot list / storyboard and the PRODUCT LOCK description for my product, and choose the target model or models — or "generic" — and each shot's input mode myself; if I then want actual video clips, take the finished prompts myself to the external model I choose, with the first-frame, first-and-last-frame or reference product images described above, arranging that provider's account and any payment myself and checking its official documentation for its current prices and limits, since this skill gives none; using the skill is a separate step that only starts when I ask for it, the skill itself sends nothing at use time, and the only use-time sends are the pastes / uploads I make to a chosen external video model, as disclosed above). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.