Content & Media Creation · prime-skills/runcomfy-agent-skills

video-edit

Name: video-edit
Author: prime-skills/runcomfy-agent-skills

Edit existing video on RunComfy — this skill is a smart router that matches the user's intent to the right edit model in the RunComfy catalog. Picks Wan 2.7 Edit-Video (general restyle / background swap / packaging swap, identity + motion preservation), Kling 2.6 Pro Motion Control (transfer precise motion from a reference video to a target character), or Lucy Edit Restyle (lightweight identity-stable restyle / outfit swap). Bundles each model's documented prompting patterns so the skill gets sharper edits without burning iterations on the wrong model. Calls `runcomfy run <vendor>/<model>/<endpoint>` through the local RunComfy CLI. Triggers on "video edit", "edit video", "restyle video", "swap video background", "motion control", "outfit swap video", or any explicit ask to transform a video.

L4 High-Privilege · Graded auth & full auditV1 Commit-PinnedExecute CommandNetwork AccessWrite FileDual-pass Audit

Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

Pure prompt routing and zero-code orchestration: The core is a 'User Intent × Video Editing Model' taxonomy table. The agent reads the table to classify the user's request into one of three models, leaving all decision-making to the host LLM. The skill file itself consists solely of the decision table, schemas, and templates.

video-edit/SKILL.md

| Restyle a talking-head video — preserve face / pose / lip movement | **Wan 2.7 Edit-Video** | Strong identity + motion preservation; supports up to 1080p |

Immediately following the routing table, the statement `The agent reads this table, classifies the user's intent, and picks the matching subsection below.` specifies the execution method—classification by the agent via the table rather than programmatic if/else logic. This skill contains no scripts; its execution runtime is the host agent.

Each of the three routes embeds a field-level schema and copy-pasteable CLI invocation templates, fully codifying the RunComfy CLI parameter protocol directly in the document so the agent can simply populate the JSON without consulting external API docs.

video-edit/SKILL.md

**Model**: `wan-ai/wan-2-7/edit-video`

Each route has a dedicated section: schema tables (prompt, video, reference_image, resolution, aspect_ratio, duration, audio_setting, seed, indicating required fields and default values) + 2 complete invocation code blocks + prompting tips. The goal is to handle requirements like 'swap background while preserving identity and original audio' in a single, well-formed JSON step.

The actual execution engine is the official RunComfy CLI and remote Model APIs: The skill indicates that the CLI handles POST requests, polling, and downloading artifacts to --output-dir, leaving only request submission and result retrieval to the local machine while all media processing occurs in the cloud.

video-edit/SKILL.md

The skill picks one of Wan 2.7 Edit-Video / Kling 2.6 Pro Motion Control / Lucy Edit Restyle based on user intent and invokes `runcomfy run <model_id>` with the matching JSON body. The CLI POSTs to the Model API, polls the request, fetches the result, and downloads any `.runcomfy.net`/`.runcomfy.com` URL into `--output-dir`. `Ctrl-C` cancels the remote request before exit.

This explains how the claimed capabilities are realized: the core generation capability resides in cloud models (Wan 2.7 / Kling 2.6 / Lucy). The skill's primary value is not generation itself, but selecting the right model and formatting prompts and parameters correctly. CLI specifics (e.g., Ctrl-C canceling remote jobs) are documented claims; the CLI source code is outside this repository.

The secondary value lies in anti-drift prompting discipline: Compressing practical knowledge on how to avoid degraded or blurry outputs into concise rules, such as prioritizing preservation targets upfront and altering only one edit direction at a time.

video-edit/SKILL.md

- **Preservation goals first**: `"Preserve [face / pose / motion / framing / lip movement]; [then state the change]"`.

The same section includes rules like `One edit direction per call. Compound edits drift on motion.`, `reference_image only when justified`, and `audio_setting: "origin" for talking-head`—all designed as countermeasures against known compound-instruction drift issues in generative models.

Hierarchical layering with brand skills in the same repository: This router handles general intent, whereas explicit mentions of a specific model are handed off to dedicated brand skills (such as wan-2-7), avoiding routing conflicts.

video-edit/SKILL.md

- **Brand-specific overrides** — if the user named a specific model, route to the corresponding brand skill (`wan-2-7`) for fuller treatment.

This structure is verified in the repository: brand directories like wan-2-7/ and kling-3-0/ coexist alongside router directories like video-edit, image-edit, and image-to-video. The router only covers generic editing intents not fully elaborated in brand-specific directories.

Security boundaries self-documented: --input is passed as a raw JSON string to the CLI (stating no shell expansion is performed), outbound endpoint whitelists, token storage location and permissions, and a 2 GiB single-file download limit.

video-edit/SKILL.md

- **Outbound endpoints**: only `model-api.runcomfy.net` (request submission) and `*.runcomfy.net` / `*.runcomfy.com` (download whitelist for generated outputs). No telemetry, no callbacks.

These declarations (including `~/.config/runcomfy/token.json` mode 0600 and the 2 GiB limit) describe CLI behavior and represent documentation-level evidence; the CLI source code cannot be independently audited from outside the repository (see verification.unlocatable).

2Core Capabilities

01Video editing intent routing: Automatically maps restyle, motion capture transfer, and outfit swap requests to Wan 2.7 Edit-Video, Kling 2.6 Pro Motion Control, or Lucy Edit Restyle

02Model-specific field-level input schemas (including required/defaults/enums/media constraints: 2–10s, ≤100MB, up to 1080p)

03Ready-to-use CLI invocation templates: complete `runcomfy run` commands with JSON body examples for each route

04Anti-drift prompting discipline guidelines (front-loading preservation goals, one edit direction per call, conditional reference media, audio preservation flags)

05Model boundary definitions: duration, resolution, direction limits, and prohibitions against combining multiple routes

06Exit code quick reference (0/64/65/69/75/77) and troubleshooting documentation links

3External Dependencies

Kind	Dependency / Endpoint
cli	runcomfy（@runcomfy/cli 官方 CLI）
package	@runcomfy/cli（npm 全局安装）及 npx skills add agentspace-so/runcomfy-skills --skill video-edit
network	RunComfy Model API（请求提交）
network	生成结果下载白名单
network	RunComfy 文档站（模型页与 CLI 故障排查，仅引用链接）
network	用户提供的第三方源媒体 URL（视频/参考图）——由 RunComfy 模型服务器抓取，非本机 CLI 抓取

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Graded/per-use authorization, sandbox recommended

For Administrators

Approval workflow, rollback support, full audit

Evidence & Triggers (File Paths & Excerpts)

网络端点: Declared external network access: model-api.runcomfy.net — Video editing request submission (the only submission channel declared in docs; via the runcomfy CLI)
本地工作目录: File read/write and artifact output: The CLI downloads generated media to the local directory specified by --output-dir (path determined by the caller)
SKILL.md: Declared invocation of host CLI interpreter or external execution environment

Audit Notes: According to the unified classification criteria: The sole runtime action is invoking the third-party npm package/CLI `@runcomfy/cli` (globally installed via npm). Each invocation reads user RunComfy credentials—either `~/.config/runcomfy/token.json` (documented as mode 0600) or the `RUNCOMFY_TOKEN` environment variable—triggering the rule: 'invokes third-party npm/CLI and potentially reads quotas/secrets → orange'. Furthermore, the execution surface (CLI, Model API, token retrieval) lies entirely outside the repository; documented CLI behaviors (0600 permissions, 2 GiB limit, absence of shell expansion, Ctrl-C cancellation) cannot be independently verified within this pin. Following the principle of 'escalating to the higher tier when uncertain', it is rated Orange. External network calls also occur: user video/reference images and prompts are sent to third-party cloud infrastructure via `model-api.runcomfy.net` (predictable target endpoint, but assets leave the boundary), reference media URLs are fetched by the server, and prompt/image injection risks are acknowledged in docs. The skill itself consists purely of markdown without scripts, obfuscation, or hidden endpoints, thus avoiding the Red tier.

5Second-Pass Verification

[ok] 三个 route 模型 ID 真实存在于调用示例与 schema — wan-ai/wan-2-7/edit-video、kling/kling-2-6/motion-control-pro、decart/lucy-edit/restyle 均在 '**Model**' 行与 `runcomfy run` 示例中成对出现，无悬空引用。
[ok] 默认路由声明一致性 — 路由表末行 'Default if unspecified | Wan 2.7 Edit-Video | Most versatile, highest resolution' 与正文默认判断一致。
[unlocatable] SKILL.md 对 CLI 行为的断言（无 shell 展开、单端点、token.json 0600、2 GiB 上限、Ctrl-C 取消远端请求） — 属作者对 @runcomfy/cli 的转述；CLI 源码不在本仓库，无法在本 pin 内独立证实，标记待补而非采信为事实。
[ok] 品牌转交声明指向 wan-2-7 — 仓库根确有 wan-2-7/ 目录；但 Kling/Lucy 无同名品牌目录（kling-3-0 为另一版本模型），brand override 仅覆盖 wan——文档未宣称覆盖 Kling/Lucy，不算矛盾。
[discrepancy] 身份/安装链接一致性 — SKILL.md 内 npx 安装命令与 GitHub 链接指向 agentspace-so/runcomfy-skills，镜像仓库 README 指向 agentspace-so/runcomfy-agent-skills，被 pin 仓库实为 prime-skills/runcomfy-agent-skills——三处来源标识不一致（非代码缺陷，属发布身份漂移，详情页应展示被 pin 的 prime-skills 仓库）。
[ok] 功能声明 vs 实际能力夸大检查 — description 所称『smart router + 提示词模式库 + runcomfy CLI 调用』全部由文档兑现；'documented prompting patterns' 中的 documented 无仓内来源（指向 RunComfy 站外文档），'sharper edits' 为无评测证据的主观宣称，已在 risks 标注。

6Verdict & Scenarios

Low decision overhead for model selection: An intent mapping table matches natural language needs (background swap, motion control, outfit swap) to the optimal cloud model, letting non-expert users get started with a single prompt

Embedded schemas and JSON templates with readily visible field constraints (duration, resolution, aspect ratio, audio), removing the need to consult external API documentation

Actionable anti-drift prompt guidelines addressing compound-instruction drift in generative models (front-loaded preservation, single-direction modifications)

Transparent security profile: Explicitly documents outbound endpoint whitelists, token storage paths and permissions, and download size limits for informed user decisions

Hierarchical coordination with brand skills: Separates generic intents from explicit model requests without conflicts, enabling smooth escalation to granular skills

Best for: Ideal for Claude/Cursor agent users with an active RunComfy account who want to delegate video background swaps, outfit changes, and motion transfer to cloud models; marketing and product teams handling batch creative iterations; and skill authors looking to reuse the 'intent routing + anti-drift discipline' prompt engineering paradigm.

Not for: Not suitable for users without a RunComfy account or those unwilling to upload video assets to a third-party cloud; local/offline editing scenarios; multi-model composite editing on the same clip (the doc explicitly restricts calls to one model at a time); scenarios requiring end-to-end auditable security (both CLI and cloud are external); or users looking to experiment with unstructured freeform iterations without adhering to the schema and discipline.

Install Agent-Ready

① Verified Mirror 2026-09-09

Option A · Direct DownloadDownload video-edit.tar.gz

sha256: 6e103a2472433792…

Option B · Copy Guide for Agent

install.json

Agents automatically download and verify mirrors from this JSON guide.

② Upstream GitHub

GitHub Repo ↗

Pinned to commit fca19ae084.