rh-workflow-9b
Skill1 starsChanged 3 months agoApache-2.0View on GitHub
9B 执行 RunningHub(RH)工作流 API 的唯一行为准则(无漂移宪章)。当你(本地 9B 模型)作为 RH 通道执行器运行时必须先读本宪章:只允许用 run_i2v_workflow.py / run_batch.py 两个入口,参数白名单,禁止写新脚本/编造参数/换主题/手写轮询,先 --check 后提交,失败不重试写 issues 上报。违反任一条 = 违规 = 停止并上报 DeepSeek。
In this repository51 other files it gives its agents.
- 3d-animation-short-generatorSkill
- brand-promo-video-generatorSkill
- colleague-cameronSkill
- colleague-muskSkill
- colleague-supermateSkill
- comfyui-zimage-9bSkill
- context-handoffSkill
- co-op-game-intro-generatorSkill
- deepseek-eyesSkill
- dot-skillSkill
- doubao-creatorSkill
- dsh-deepseek-vision-bridgeSkill
- dsh-im-wecomSkill
- cordis-plugin-developmentSkill
- dsh-archive-agent-notesSkill
- dsh-code-reviewSkill
- dsh-doc-site-syncSkill
- dsh-doc-standardsSkill
- dsh-find-simplificationsSkill
- dsh-merging-stacked-prsSkill
- dsh-pre-push-checksSkill
- dsh-prose-standardSkill
- dsh-translate-docsSkill
- dsh-trim-cot-leakageSkill
- editing-cordis-compositionsSkill
- record-browser-gifSkill
- dsh-upgrade-runbookSkill
- h3-prompt-writingSkill
- handdrawn-live-video-generatorSkill
- hermes-memorySkill
- image-analysisSkill
- lyric-mv-storyboardSkill
- lyric-to-storyboard-face-lockSkill
- lyric-to-storyboard-lrcSkill
- lyric-to-storyboardSkill
- minimalist-product-ad-generatorSkill
- h3-video-producerSkill
- modelscope-apiSkill
- multi-agent-orchestratorSkill
- music-video-subtitle-generatorSkill
- paper-collage-explainer-generatorSkill
- papercraft-stop-motion-explainerSkill
- quark-driveSkill
- quark-qwen-visionSkill
- rh-workflowSkill
- supermate-code-reviewSkill
- Supermate-harness-launcherSkill
- supermate-scriptSkill
- colleague-supermateSkill
- video-production-studioSkill
- visual-file-organizerSkill
The file
Files
9B RH 执行规范(无漂移宪章)
核心:RH API 是最简单的机械程序,不需要创造,只需要照做。
1. 唯一入口(只准这两个,禁止写新脚本)
cd "<RH_Workflow_Api 脚本目录>" # 台式机示例:E:\Harness Workspace\Local_LLM\RH_Workflow_Api
python run_i2v_workflow.py ... # 单任务
python batch\run_batch.py --csv ... # 批量2. 参数白名单(run_i2v_workflow.py)
--image --prompt / --prompt-file --duration --resize-max 1280(必带) --output-dir --json --check --image-name
不存在:--submit --task_id --first_frame_path 等——argparse 没有就是没有。
3. 流程铁律
- 先
--check(免费自检) --resize-max 1280:参考图绝不裁剪只缩放- 提交(= 花钱,参数核对三遍)
- 轮询/下载交给脚本或
run_workflow_api.RunningHubClient——禁止手写轮询(有 float 格式化崩溃前科)
4. 提示词铁律
开头必写:第0帧起画面即为动态场景,禁止参考图版面/三视图/多视角面板/文字/白底站姿停留,人物从第一帧起连续运动。结构:人物锁定→场景→动作(进行时)→运镜→风格。
5. 失败协议
报错→不重试→写 batch\issues\<id>.md→等 DeepSeek 诊断。一次性偶发(VRAM)最多重试 1 次。
6. 状态查询
只信 API(POST /openapi/v2/query);网页工作台滞后属正常,不误判卡住。
7. 漂移违规清单(命中即停)
自己写新脚本 / 编造参数 / 换主题(橘猫事件)/ 改 Linux 路径 / 手写轮询 / 乱加 usePersonalQueue / 跳过 --check。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

