360 Secure SkillHub
360 安全技能中心

Community Site One-Shot Clips

该 Skill 将授权社区活动现场照片按拍摄顺序逐张生成 2-15 秒静音短片,不拼接;先生成标注式单片段计划,读取 image_to_video 模型与实时价格,经用户确认后通过包内 MCP 客户端提交并轮询任务,最后按顺序交付片段、实际尺寸、时长和扣费信息。

beatra-ai
beatra-ai
热度 19.65
最近更新 2026-09-06 21:23:25
分类:创作依赖:代码执行 / 读写文件 / MCP 服务应用场景:图像生成 / 视频编辑 / 工作流

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add community-site-clip
Metadata
name
community-site-clip
description
Turn authorized community event site photos and office-supplied facts into one community site clip per photo. This event site clip studio makes a silent 2-15s site photo clip from each field still clip. Use it for community event clip sets that stay one photo one clip.

Community Site One-Shot Clips

Turn authorized community event site photos and office-supplied facts into one short clip per photo. Deliver the clips in photo order. Do not stitch them.

Scope and adjacent routes

Use this Skill when a street office wants one short motion clip for each event-site photo they already captured.

Route a talking recap from one photo plus a script to wrong-item-talking. Route assembly stills to assembly-one-step-clip. Route a generic still that is not a labeled event site to image-to-motion. Route classroom boards to blackboard-one-shot. Do not look up public posts for missing event facts.

Inputs and defaults

The hard inputs are:

  • at least one accessible event-site photo the host Agent can inspect, in capture order;
  • office-supplied facts that name what each photo shows (event, location, moment, or sequence number).

Do not invent attendance, a ranking, an official stamp, or a spoken line. A photo is not a source for unstated wording. Reuse event name, language, destination, and must-keeps already in the conversation. Ask only for a missing hard input.

Default to one beatra.videos.animate call per photo, model: "auto", the photo as the strict first frame, no driving_audio, and a source-derived aspect ratio. Choose a supported integer duration in the 2-15s range with the user; if they do not pick, use 5s.

Golden path

  1. Inspect every photo. Record MIME type, width, height, aspect ratio, byte size, and whether it has an alpha channel. For a local file, upload only through the bundled client after inspection. Never pass a local path to a remote tool.

  2. Write a labeled one-clip plan from the office's facts and what the photos actually show. Keep capture order. Leave a photo whose fact is missing on the plan. That plan is the free visible result. Planning is not approval.

  3. Call beatra.models.list with {"capability":"image_to_video"}. Admit each photo against a current card that accepts a first-frame image without driving audio. Compare every image fact and the chosen duration with advertised constraints. If any required media fact is unavailable or incompatible, stop before video.

  4. Show one current production card and wait before any video client_request_id or beatra.videos.animate call:

    1. Work — one community-site clip per named photo, a generate call (beatra.videos.animate).
    2. Credits — the live image_to_video price just read. Do not reuse a remembered number.
    3. Count — one paid call per photo. Do not batch photos into one submit.
    4. Identity — one new opaque client_request_id per photo. A changed prompt, photo, duration, model, or control mints a new ID.
    5. If we stop here — the labeled clip plan remains usable.
    6. If the balance is insufficient — relay the official message and its top-up URL exactly. Translate the prose; keep the URL. Do not retry until the office says they have topped up.

    Do not submit until the user confirms they have topped up or already have enough credits.

  5. Submit beatra.videos.animate exactly once per admitted photo through bundled scripts/mcp_client.py. Do not configure a host Beatra Connector. Do not use REST/OpenAPI as a fallback. Poll each video task with beatra.tasks.get until terminal.

  6. Deliver the clips in photo order with actual dimensions, duration, usage, and billing.net_charged_credits. Review site identity, motion, and must-keep drift. Never invent a stitch, concat, or editor tool.

Decisions that require confirmation

Planning and live price quotes are free. They are not approval. Video animate needs the six-field card above. File access is not consent to reuse a photo outside this event set.

Recovery

Each paid photo has its own frozen payload and ID. Recover a lost create response only with that photo's identical payload. Recover a lost task ID through beatra.tasks.list and beatra.tasks.get. Call beatra.tasks.cancel only when the user asks to cancel that photo; on 409 keep polling. On insufficient_balance, keep the top-up URL inside the balance error exact and retry the same frozen ID only after the user says they have topped up. If a create returns no task_id, do not poll; reconcile before minting a new ID. A changed photo, prompt, duration, or model is new work and a new card.

Execution

Invoke every remote Beatra operation only through this package's bundled scripts/mcp_client.py. Put the MCP tool name after call and send one JSON object on standard input.

python3 scripts/mcp_client.py call beatra.models.list
{"capability": "image_to_video"}

Do not configure or call a host Beatra Connector, and do not use REST/OpenAPI as a fallback.

Account balance

When the user asks how many credits remain or whether a live estimate fits, call beatra.wallet.get. When they ask what was charged, call beatra.wallet.ledger. Both are read-only. Do not invent an account-balance or top-up tool. Do not make wallet.get a required step before every paid submit.

When a model card comes back carrying a top_up block, relay its tiers as the card lists them and in that order. Do not rank them, do not talk one down, and do not pick one for the user. Which tier suits them is their call, made on the wallet page with the whole list in front of them. Never quote a tier from memory.

References by task

Runtime and safe automatic updates

The bundled client silently checks for a newer release at most once every 24 hours per installation. When a newer version is available, it installs automatically without separate confirmation. It downloads only from the fixed official Beatra discovery and immutable CDN paths for this package, channel, and locale, verifies discovery data, archive, manifest, and every packaged file, and replaces only package-owned files.

Update checks, downloads, verification, replacement, rollback, and recovery fail open: the current installation remains usable and the original command continues. An update failure never authorizes retrying a paid video request. The setting persists for this installation. See automatic updates and safety.

python3 scripts/mcp_client.py update --auto off
python3 scripts/mcp_client.py update --auto on
python3 scripts/mcp_client.py update --check