Nbcraft

What it does

When your agent uses it

• User wants to generate images
• Run batch image generation jobs (Gemini only)
• Reverse-analyze images (UI structure / general description
• Check generation history/stats

Example prompts

• “nano banana”
• “Gemini 生图”
• “seedream”
• “/nbcraft”

Requirements

• Python 3
• A credential in GEMINI_API_KEY
• A credential in DASHSCOPE_API_KEY

Workflow steps

7 steps, taken from the step headings in SKILL.md.

1. 1Generate — nb gen
2. 2Batch API — nb batch (Gemini only)
3. 3Reverse Analysis — nb rev (Gemini only)
4. 4Video — nb video (Seedance 2.0 on Volcengine Ark)
5. 5History — nb history
6. 6Config — nb config
7. 7Stats — nb stats

What it can do on your machine

Read from SKILL.md and the folder at commit 21fc7a3. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  Ships script files (Python, from the files we listed), which the agent can run.

  Shell commands in SKILL.md call:

  • pip

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  Hosts in commands or code, which the agent is likely to contact:

  • api.openai.com

  Also links to:

  • console.volcengine.com

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names these keys or tokens, usually read from environment variables:

  • GEMINI_API_KEY
  • DASHSCOPE_API_KEY
  • ARK_API_KEY
  • OPENAI_API_KEY

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Nbcraft loads about 2.8k tokens when it runs. Until then it costs about 248 tokens; SKILL.md has 754 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~248

When it runs · the whole SKILL.md, loaded when a task matches

~2.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from jieyefriic/nbcraft at commit 21fc7a3, republished under its MIT licence (© jieyefriic). 754 words, ~2,846 tokens.

name

nbcraft

description

Multi-backend Image + Video Generation CLI (nb command). Image: Gemini (pro=gemini-3-pro-image-preview, flash=gemini-3.1-flash-image-preview), DashScope 阿里云百炼 (wan-pro=wan2.7-image-pro, wan=wan2.7-image), Volcengine Ark 火山方舟 字节 Seedream (seedream=doubao-seedream-4-5-251128, seedream-lite=doubao-seedream-5.0-lite, seedream-legacy=doubao-seedream-4-0-250828), OpenAI (gpt=gpt-image-2). Video: Volcengine Ark Seedance 2.0 (seedance=doubao-seedance-2-0-260128, seedance-fast=doubao-seedance-2-0-fast-260128). Use when user wants to generate images or videos, run batch image generation jobs (Gemini only), reverse-analyze images (UI structure / general description, Gemini only), check generation history/stats, or manage API configuration. Triggers on "nb", "nano banana", "生成图片", "出图", "Gemini 生图", "万相", "wan", "百炼出图", "seedream", "字节出图", "豆包出图", "火山方舟", "ark", "gpt-image", "gpt 出图", "openai 出图", "生成视频", "出视频", "seedance", "视频生成", "batch 生成", "分析图片", "逆向", "UI结构", "describe image".

# Basic (Gemini pro)
nb gen -p "描述"

# Gemini with reference image + aspect ratio + resolution
nb gen -p "描述" -i ref.png --ar 3:4 --res 2K -o output.png

# 万相 wan2.7-image-pro 4K 文生图
nb gen -p "描述" --model wan-pro --ar 16:9 --res 4K

# 万相带参考图做图像编辑（最多 9 张）
nb gen -p "把这只狗换成宇航员" -i dog.png --model wan --ar 1:1 --res 1K

# Seedream 4.5 文字渲染（海报/PPT/电商主图）
nb gen -p "中文海报，主标题『春日上新』，副标题小字" --model seedream --ar 3:4 --res 4K

# Seedream 多参考图融合（最多 14 张）
nb gen -p "保留参考图1的服装，融合参考图2的姿势" -i a.png -i b.png --model seedream

# Seedream 5.0 Lite，更快更便宜
nb gen -p "概念图" --model seedream-lite --ar 16:9 --res 2K

# GPT Image 2 文字渲染 + agentic 推理
nb gen -p "极简 logo 草图，标语 'Stay Curious'" --model gpt --ar 1:1 --res 2K

# GPT Image 2 编辑（自动走 /v1/images/edits）
nb gen -p "把背景换成森林" -i photo.png --model gpt

# DashScope 负向提示词
nb gen -p "描述" --model wan --negative-prompt "低分辨率, 畸形, 模糊"

# Prompt from file
nb gen -p @prompt.txt

# Repeat same prompt N times (1-8), outputs auto-numbered
nb gen -p "描述" --repeat 4 -o out.png
# → out_1.png, out_2.png, out_3.png, out_4.png

# Multiple reference images
nb gen -p "描述" -i a.png -i b.png -i c.png

# Submit and wait for results
nb batch submit -f tasks.jsonl --poll -o ./output/

# Submit without waiting; pick model
nb batch submit -f tasks.jsonl --name "my-job" --model flash

# Check status
nb batch status <job_name>

# Download completed results
nb batch download <job_name> -o ./output/

# List all jobs
nb batch list

{"key": "img1", "prompt": "描述", "images": ["ref.png"], "ar": "3:4", "res": "2K"}
{"key": "img2", "prompt": "另一个描述", "ar": "1:1"}

# 分析 UI 截图的 HTML 结构
nb rev struct -i screenshot.png

# 保存结果到文件
nb rev struct -i screenshot.png -o structure.md

# 追加额外指令
nb rev struct -i screenshot.png --extra "重点分析导航栏部分"

# 默认详细描述（中文）
nb rev desc -i photo.png

# 自定义分析 prompt
nb rev desc -i photo.png -p "分析这张图的配色方案"

# 保存结果
nb rev desc -i photo.png -o analysis.txt

# 文生视频（默认 seedance、720p、5s、无声）
nb video gen -p "一只柴犬在樱花树下慢镜头转身" --ar 16:9

# 图生视频（首帧/参考图）
nb video gen -p "镜头从产品平移到背景" -i product.png --duration 8

# 高质量 1080p + 同步音频 + 自动下载
nb video gen -p "电影级人物对白镜头" --model seedance --res 1080p --duration 10 --audio -o out.mp4

# Fast 版快速迭代
nb video gen -p "概念草图" --model seedance-fast --duration 4

# 异步：只提交，不等
nb video gen -p "..." --no-poll

# 查询任务状态
nb video status <task_id>

# 单独下载已完成的视频（task_id 必须 status=succeeded）
nb video download <task_id> -o out.mp4

nb history                    # Recent 20 records
nb history -n 50              # Last 50
nb history -s "keyword"       # Search by prompt
nb history <record_id>        # Full detail (JSON)

nb config show
nb config set api_key <key>            # Gemini
nb config set dashscope_api_key <key>  # DashScope 国内
nb config set ark_api_key <key>        # Volcengine Ark 火山方舟
nb config set ark_endpoint_id ep-xxx   # 可选，传了就用 endpoint
nb config set openai_api_key <key>     # OpenAI
nb config set openai_base_url <url>    # 可选，自定义 base URL / 代理
nb config set default_ar 3:4
nb config set default_res 2K
nb config set default_model flash       # pro | flash | wan-pro | wan | seedream | seedream-lite | seedream-legacy | gpt
nb config set poll_interval 20          # Batch poll seconds
nb config set output_dir ./out          # Default output directory

nb stats    # Total calls, success/fail, direct/batch, monthly/daily breakdown

© jieyefriic, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 24 other files in the repository root of jieyefriic/nbcraft.

• SKILL.md
• .github/ISSUE_TEMPLATE/bug_report.md
• .github/ISSUE_TEMPLATE/feature_request.md
• .github/PULL_REQUEST_TEMPLATE.md
• .github/workflows/ci.yml
• .gitignore
• CHANGELOG.md
• CODE_OF_CONDUCT.md
• CONTRIBUTING.md
• LICENSE
• README.md
• SECURITY.md
• nbcraft/__init__.py
• nbcraft/__main__.py
• nbcraft/cli.py
• nbcraft/client.py
• nbcraft/store.py
• … and 8 more

Open the folder on GitHubat commit 21fc7a3