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

happyhorse-1-0

Name: happyhorse-1-0
Author: prime-skills/runcomfy-agent-skills

Generate text-to-video with HappyHorse 1.0 on RunComfy. Documents HappyHorse 1.0's strengths (#1 on Artificial Analysis Video Arena, native 1080p with in-pass synchronized audio, multi-shot character consistency, 6-language prompt support), the duration / aspect-ratio / resolution schema, and when to route to Wan 2.7 / Seedance 2 / LTX 2 instead. Calls `runcomfy run happyhorse/happyhorse-1-0/text-to-video` through the local RunComfy CLI. Triggers on "happyhorse", "happy horse", "happyhorse 1.0", "happyhorse video", or any explicit ask to generate video with this model.

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

Implemented via a 'zero-code + external CLI delegation' approach: `happyhorse-1-0/` contains only a single `SKILL.md` with no scripts. At runtime, the host agent executes `runcomfy run happyhorse/happyhorse-1-0/text-to-video --input '<JSON>' --output-dir <path>` via Bash. Both video and audio are generated on the RunComfy cloud; the skill is solely responsible for translating intent into proper JSON inputs and prompts.

happyhorse-1-0/SKILL.md

1. The skill invokes `runcomfy run happyhorse/happyhorse-1-0/text-to-video` with a JSON body matching the schema.

Verified with `git ls-tree HEAD` that `happyhorse-1-0/` contains only `SKILL.md` as an entry. This skill is the only one among the five that explicitly breaks down the invocation lifecycle into 5 numbered steps (the other four contain only a single prose paragraph).

Complete 5-step lifecycle protocol: CLI sends a POST request with a bearer token to the Model API → receives a `request_id` → polls GET `.../requests/<id>/status` every 2 seconds → upon reaching terminal status, calls GET `.../requests/<id>/result` → downloads URLs only if the host ends with `.runcomfy.net` or `.runcomfy.com`, merely listing other URLs without fetching them.

happyhorse-1-0/SKILL.md

4. On terminal status, the CLI fetches `GET .../requests/<id>/result` and downloads any URL whose host ends with `.runcomfy.net` or `.runcomfy.com` into `--output-dir`. Other URLs are listed but not fetched.

'Download only whitelisted hosts, list others without fetching' is the most explicit data retrieval discipline specified across this family. Combined with step 5 ('POST .../cancel on Ctrl-C to avoid billing for stopped GPUs'), it constitutes the unique cost-control specification of this skill.

The input schema contains 7 fields covering text, aspect ratio, resolution, duration, reproducibility, and watermark: `prompt` (≤2500 characters, 6 languages), `aspect_ratio` (5 values, default 16:9), `resolution` (720P/1080P, default 1080P), `duration` (3–15 seconds, default 5), `seed` (default 0), and `watermark` (default true).

happyhorse-1-0/SKILL.md

| `watermark` | bool | no | true | Provider watermark. |

A toggleable `watermark` is unique to this skill within the family (defaults to true to retain the provider watermark). The default 1080P resolution is also the highest default across the five skills, aligning with its 'native 1080p' positioning.

Prompting techniques target the temporal dimension: describe action progressions rather than static frames (chaining multiple temporal verbs), front-load camera shots and angles, and structure iterations as one visual beat per segment.

happyhorse-1-0/SKILL.md

**Describe motion over time, not a still.** "A woman turns from the window, walks two paces to the desk, picks up the cup, lifts it to her face, takes a sip" beats "a woman drinking coffee".

This explains the mechanism behind 'why static frame descriptions turn blurry' (lack of temporal verbs leaves motion priors with no anchoring point)—one of the few documentation sections explaining 'why' instead of just stating 'how'.

Multi-shot consistency is achieved by 're-stating anchors in every shot': when writing two beats, the same clothing/props (e.g., 'red wool coat / blue scarf') must be reiterated in each shot for the model to maintain cross-shot character identity.

happyhorse-1-0/SKILL.md

**Multi-shot consistency** — when describing two beats, restate the anchor at each: `"Shot 1: tall woman in red wool coat, blue scarf, in a rainy alley. Shot 2: same woman in red coat / blue scarf, now ducking under an awning."` HappyHorse holds the look but needs the anchor.

Clarifying that 'the model can maintain identity, but requires anchors' lays out both the capability and its prerequisite, preventing users from assuming cross-shot consistency happens automatically.

Audio is generated in the same pass (dialogue, ambient, Foley) but does not accept external audio tracks to drive lip-sync. For audio-driven workflows, users should switch to Wan 2.7 (`audio_url`) or Seedance 2.0 Pro—a boundary explicitly documented under Limitations.

happyhorse-1-0/SKILL.md

- **Audio is in-pass only** — you can't pass external audio to drive lip-sync. For audio-driven lip-sync, use Wan 2.7 (which accepts an `audio_url`) or Seedance 2.0 Pro.

Establishes a clear division of labor with `seedance-v2`: this skill generates audio (controllable tone), while `wan-2-7` ingests audio (aligning lip-sync to user audio tracks).

Self-limiting section (unique within the family): explicitly clarifies that it is 'not a self-hosted video runner' and 'not a capability grant; relies on an available RunComfy account', clearly pinning down the skill's scope.

happyhorse-1-0/SKILL.md

Not a self-hosted video runner. Not a capability grant — depends on a working RunComfy account.

Only this skill among the five features a '## What this skill is not' section, preempting the common misconception that 'a skill equals local capability'.

The routing table maps out the roles of other family video models (audio-driven lip-sync → Wan 2.7; reference video + fine dialogue lip-sync → Seedance 2.0 Pro; motion editing → Kling; fast iteration → LTX 2), while declaring that explicit user requests route back to this skill unconditionally.

happyhorse-1-0/SKILL.md

| Detailed lip-synced dialogue + reference video | Seedance 2.0 Pro |

Matches the document note: 'If the user said "HappyHorse" / "happy horse video" explicitly, route here regardless.' Wan 2.7 and Seedance 2.0 have matching skill directories in the repository, whereas Kling Video O1 and LTX 2 do not (see second_pass).

2Core Capabilities

01Text-to-video generation (single t2v endpoint, prompt ≤ 2500 characters)

02Native 1080p output (720P / 1080P selectable, default 1080P)

03In-pass synchronized audio (dialogue, ambient, Foley), specifying what is heard via prompt

04Multi-shot character and costume consistency (via re-stating anchors per shot)

056-language prompt support (CN, EN, JP, KR, DE, FR)

06Duration and aspect ratio control (3–15 seconds; 5 aspect ratios, non-enum values return 422)

07Provider watermark toggle (`watermark` defaults to true, can be set to false)

08Documented failure surfaces and cost control (sysexits return codes + Ctrl-C remote cancellation to avoid billing for stopped GPUs)

3External Dependencies

Kind	Dependency / Endpoint
cli	runcomfy（RunComfy 官方 CLI）
package	@runcomfy/cli（npm 全局包）
package	npx skills（skills.sh 安装器）
api	RunComfy Model API（HappyHorse 1.0 text-to-video）
api	RunComfy 请求状态/结果/取消端点（轮询与中断）
network	RunComfy 产物下载白名单
network	RunComfy 站点/模型页/文档（文档引用，非运行时调用）
network	GitHub（技能源码链接）

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)

scripts/: Local script or command execution: (No scripts inside the skill itself) Host agent executes external CLI via Bash: `runcomfy run <vendor>/<model>/<endp`
网络端点: Declared external network access: POST `https://model-api.runcomfy.net/v1/models/happyhorse/happyhorse-1-0/tex`
本地工作目录: File I/O and artifact output: Generated outputs are downloaded and written to `--output-dir` by the external CLI (documentation notes abortion if a single file exceeds 2 GiB to prevent disk exhaustion)

Audit Notes: Per the unified grading standard for this review (aligned with orange for image-edit in the same repo and other skills in this batch): at runtime, it invokes the third-party npm package/CLI `@runcomfy/cli`, which reads user RunComfy credentials (`~/.config/runcomfy/token.json` or `RUNCOMFY_TOKEN`) on every execution—matching the 'invoking third-party CLI/remote package accessing credentials/quotas' tier. Claims regarding CLI implementation, Model API behavior, 0600 token permissions, 2 GiB limits, exit code semantics, and step 5 cancel HTTP verbs all reside outside the repository without source code to verify in this pin; according to 'when in doubt, bump to the higher tier', orange is assigned. Additionally, 1080p video represents high-unit-cost inference billed per invocation, which when combined with exit code 75 retry logic may amplify credit consumption. Generated artifacts are videos with synchronized audio and optional watermarks, presenting deepfake-related abuse surfaces (not warned in documentation). The skill itself is purely Markdown with no scripts, obfuscation, or hidden endpoints, thus avoiding the red tier.

5Second-Pass Verification

[ok] skill 目录内容 = 仅 SKILL.md — git ls-tree HEAD 与 ls -a 双向核实；仓库根仅 LICENSE + README.md + 31 个技能目录，无 scripts/、plugins/、.claude/。
[ok] 全部外部依赖条目回查调用点真实存在 — runcomfy CLI 调用点 3 处（默认/竖屏/720p 试跑）、npm 安装 1 处、npx skills 1 处、model-api POST 1 处、status/result/cancel 三个端点各 1 处、白名单 1 处、文档/模型页/GitHub 链接多处，均实读命中。
[discrepancy] Ctrl-C 取消的 HTTP 语义与同仓 CLI 说明书一致 — 本技能写『Ctrl-C 时发送 POST .../requests/<id>/cancel』，同仓 runcomfy-cli/SKILL.md 写『Ctrl-C sends DELETE to the request endpoint』，其余四份兄弟技能只写 '`Ctrl-C` cancels the remote request before exit.'（不含动词）。同一个 CLI 在两份文档里被描述成两种 HTTP 动词，至少一处不准确；CLI 源码不在仓内，无法在本仓判定孰是。
[unlocatable] #1 on Artificial Analysis Video Arena（Elo 1333 t2v / 1392 i2v）主张 — 该排名出现在 description、正文首段与 Where it shines 三处，但本技能未给任何链接或数据来源；同仓 ai-video-generation/SKILL.md 与 image-to-video/SKILL.md 复述同一主张（后者写 Elo 1392）。榜单本身在仓外（artificialanalysis.ai），本次未做独立核实，记为源码内无法定位的证据。
[discrepancy] 路由建议的目标模型在本仓有对应技能 — 'Cinematic motion editing on existing footage | Kling Video O1' 与 'Ultra-fast iteration (sub-second per frame) | LTX 2' 在仓库内没有同名技能目录（全仓无 ltx-* 目录；kling-3-0/SKILL.md 自称 'Kling 3.0 (also called Kling V3.0)'）。Wan 2.7 / Seedance 2.0 Pro 两条有同名目录，属有效引用。
[unlocatable] Security & Privacy 的技术断言可在本仓证实 — 0600 token 权限、2 GiB 下载上限、无 shell 展开、下载白名单、退出码语义、轮询间隔 2s 与取消动词均为仓外 @runcomfy/cli 行为；仓内无 CLI 源码或版本 pin，只能作为文档自述引用。
[discrepancy] 安装一行与 GitHub 链接指向旧仓 agentspace-so/runcomfy-skills — 安装行 `npx skills add agentspace-so/runcomfy-skills --skill happyhorse-1-0 -g` 与 GitHub 链接 github.com/agentspace-so/runcomfy-skills 指向旧仓；本仓 pin 在 prime-skills/runcomfy-agent-skills（README 用 runcomfy-agent-skills）。gh api：旧仓仍存在（5★、pushed 2026-04-29、含 happyhorse-1-0 目录）；prime-skills/runcomfy-agent-skills 与 agentspace-so/runcomfy-agent-skills 为同一仓（相同 stats）。按文档安装可能装到旧版。
[discrepancy] 『watermark 可关』的安全含义是否被文档提示 — schema 给出 `watermark` 默认 true、可设 false（示例小节标题即 '**Vertical short (9:16, 8s, no watermark):**' 并把 watermark 设为 false），但 Security & Privacy 段完全未提『关闭供应方水印后产物无来源标识』这一滥用面（本族其它技能不提供该开关，无同类提示可比）。属安全提示覆盖不全，非代码风险。

6Verdict & Scenarios

Most concrete lifecycle and cost-control documentation: 5 numbered steps (submit → `request_id` → poll every 2s → fetch result → whitelist-only download) and explains that Ctrl-C cancels remote tasks to prevent billing on stopped GPUs.

Highest default delivery specification: defaults to 1080P with vendor watermark retained, suitable for commercial delivery; also provides a 720P/3s 'cheap trial run' mode to manage costs.

Provides actionable practices for multi-shot consistency (re-stating clothing/prop anchors per shot) while honestly acknowledging that 'the model can maintain identity, but requires anchors'.

Explains the mechanism behind prompting techniques rather than just listing rules (static frame descriptions cause motion blur; use temporal verbs instead), complete with three copy-pasteable example prompts.

Clear self-boundaries (not a self-hosted runner, not an entitlement grant), avoiding the misconception that the skill provides local compute or licensing.

Best for: Ideal for: brand videos and product demos requiring native 1080p and in-pass synchronized audio (dialogue/ambient/Foley); multi-shot narratives demanding cross-shot character/wardrobe consistency (following the 're-stating anchors per shot' pattern); multilingual short ads (6 prompt languages); and cost-conscious pipelines that run a 720P/3s preview before final delivery. Prerequisite: RunComfy CLI installed and logged in (or `RUNCOMFY_TOKEN` configured).

Not for: Not suitable for: driving precise lip-sync with custom audio tracks (use `wan-2-7`'s `audio_url` or Seedance 2.0 Pro instead); single continuous shots >15s (must be broken into segments); image-to-video (documentation clarifies this t2v template is strictly text-driven); ultrawide aspect ratios (supports only 5 aspect ratios; ultrawide inputs will be cropped or rejected); or compliance environments requiring data/assets to remain air-gapped or code execution to be fully auditable—all operations for this skill occur via an external CLI on the RunComfy cloud.

Install Agent-Ready

① Verified Mirror 2026-09-15

Option A · Direct DownloadDownload happyhorse-1-0.tar.gz

sha256: ed6d01d2aaa14890…

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.