draw-thing

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Draw Thing

Draw Thing

Local AI image and media generation through
draw-things-cli
. Use the current Draw Things command surface:
generate
,
models list
,
models ensure
,
models import
, and
train lora
.
Scope: Local Draw Things image generation, image editing, model setup, model inventory, imports, and short local media generation. NOT for UI implementation (design), ad creative strategy (ad-creative), cloud image APIs, or general AI-model market research.

通过
draw-things-cli
实现本地AI图像与媒体生成。使用当前Draw Things的命令集:
generate
models list
models ensure
models import
train lora
适用范围: 本地Draw Things图像生成、图像编辑、模型设置、模型清单管理、模型导入以及本地短媒体生成。不适用于UI实现(设计)、广告创意策略(广告创意)、云端图像API调用或通用AI模型市场调研。

Canonical Vocabulary

标准术语表

TermMeaningNOT
txt2imgText-to-image generation: prompt in, image file outimg2img
img2imgImage-to-image generation: input image plus prompttxt2img
mediaVideo or audio-video generation with
--frames
and video output
UI animation
model idDraw Things model filename or model reference accepted by
--model
display name guess
recommended packCurrent rig-aware model set approved for this Macevery installed model
legacy modelOlder model kept for compatibility, not recommended by defaultdeletion target
cfgCurrent CLI guidance flag:
--cfg
old
--guidance-scale
strengthImg2img change amount: lower preserves, higher redrawsoutput scale
seedReproducibility value for identical prompts/settingsmodel selection
config fileJSON file passed with
--config-file
for advanced app settings
direct CLI flag

术语含义排除项
txt2img文本转图像生成:输入提示词,输出图像文件img2img
img2img图像转图像生成:输入图像加提示词txt2img
media使用
--frames
参数生成视频或音视频内容
UI动画
model idDraw Things模型文件名或
--model
参数接受的模型引用标识
猜测的显示名称
recommended pack当前针对该Mac设备配置的推荐模型集合所有已安装模型
legacy model为兼容保留的旧模型,默认不推荐使用待删除目标
cfg当前CLI的引导参数:
--cfg
旧参数
--guidance-scale
strengthImg2img的修改幅度:值越小保留原图越多,值越大重绘程度越高输出尺寸
seed用于确保相同提示词/设置生成相同结果的可复现值模型选择
config file通过
--config-file
传入的高级应用设置JSON文件
直接CLI参数

Dispatch

命令调度

$ARGUMENTS
ModeAction
generate <prompt>
/
create <prompt>
Generatetxt2img with best current image defaults
Natural language image descriptionAuto: GenerateDetect prompt intent and choose a model
edit <path> <prompt>
/
transform <path>
Editimg2img with
--image
and
--strength
Path to image + modification intentAuto: EditPreserve original, write a new output file
video <prompt>
/
media <prompt>
/
animate <prompt>
MediaGenerate short local media with
--frames
setup
/
download models
/
best models
SetupInventory rig and ensure recommended pack
inventory
/
models
/
model status
InventoryRun model inventory helper and summarize gaps
import <path-or-url>
ImportUse
draw-things-cli models import
model <name>
Model infoShow current role, settings, and status
batch <prompt>
/
variations <prompt>
BatchUse a shell loop with multiple seeds and outputs
refine
/
iterate
RefineRe-run with a locked seed and one changed parameter
upscale <path>
/
controlnet ...
/
lora ...
AdvancedVerify config-file workflow before running
(empty)HelpVerify CLI, show modes and examples
$ARGUMENTS
模式操作说明
generate <prompt>
/
create <prompt>
生成使用当前最佳图像默认参数执行txt2img
自然语言图像描述自动:生成检测提示词意图并选择对应模型
edit <path> <prompt>
/
transform <path>
编辑使用
--image
--strength
参数执行img2img
图像路径+修改意图自动:编辑保留原图,生成新的输出文件
video <prompt>
/
media <prompt>
/
animate <prompt>
媒体生成使用
--frames
参数生成本地短媒体内容
setup
/
download models
/
best models
设置检测设备配置并确保推荐模型集合已安装
inventory
/
models
/
model status
清单管理运行模型清单助手并汇总缺失的模型
import <path-or-url>
导入使用
draw-things-cli models import
命令
model <name>
模型信息显示模型当前角色、设置和状态
batch <prompt>
/
variations <prompt>
批量生成使用Shell循环搭配多个seed生成多份输出
refine
/
iterate
优化锁定seed,仅修改一个参数重新运行
upscale <path>
/
controlnet ...
/
lora ...
高级操作运行前先验证配置文件工作流
(空参数)帮助验证CLI,显示模式及示例

Auto-Detection Heuristic

自动检测规则

  1. Prompt asks to install, download, configure, or choose best models -> Setup.
  2. Prompt asks for model list, current downloads, or missing models -> Inventory.
  3. Prompt asks for video, motion, mp4, mov, frames, or animation -> Media.
  4. Existing file path plus edit/restyle/change/transform -> Edit.
  5. Existing file path plus upscale/control/lora/inpaint -> Advanced, then verify current CLI/config support before running.
  6. Descriptive text with no file path -> Generate.
  7. Ambiguous -> ask which mode and state the tradeoff.

  1. 提示词涉及安装、下载、配置或选择最佳模型 → 设置模式。
  2. 提示词询问模型列表、当前下载情况或缺失模型 → 清单管理模式。
  3. 提示词涉及视频、动态、mp4、mov、frames或动画 → 媒体生成模式。
  4. 包含现有文件路径加编辑/重新风格化/修改/转换需求 → 编辑模式。
  5. 包含现有文件路径加上 upscale/control/lora/inpaint 需求 → 高级操作模式,运行前需验证当前CLI/配置支持情况。
  6. 仅包含描述性文本且无文件路径 → 生成模式。
  7. 意图模糊 → 询问用户选择哪种模式并说明权衡点。

Prerequisite Protocol

前置操作流程

Run before any generation or model mutation:
  1. Check CLI:
    command -v draw-things-cli
    .
  2. If missing, show install command and stop:
    bash
    brew install drawthingsai/draw-things/draw-things-cli
  3. Verify installed command surface:
    draw-things-cli generate --help
    .
  4. For setup/inventory, prefer:
    bash
    uv run python skills/draw-thing/scripts/model_inventory.py --format json
  5. Always pass
    --output
    ; without it, the current CLI may only preview in supported terminals instead of writing a file.

在执行任何生成或模型变更操作前需完成以下步骤:
  1. 检查CLI是否存在:
    command -v draw-things-cli
  2. 若未安装,显示安装命令并终止操作:
    bash
    brew install drawthingsai/draw-things/draw-things-cli
  3. 验证已安装的命令集:
    draw-things-cli generate --help
  4. 对于设置/清单管理操作,优先使用:
    bash
    uv run python skills/draw-thing/scripts/model_inventory.py --format json
  5. 必须传入
    --output
    参数;若省略,当前CLI可能仅在支持的终端中预览图像,而非写入文件。

Current Best Model Policy

当前最佳模型策略

As of 2026-05-15, this Mac is Apple M1 Pro, arm64, macOS 26.5, 32 GiB unified memory. Use current Draw Things optimized checkpoints and do not recommend legacy SDXL/SD1.5/FLUX.1 families unless the user explicitly asks for compatibility.
RoleModel idStatusDefault use
Fast image default
z_image_turbo_1.0_q8p.ckpt
recommendedFast drafts, 8 steps,
--cfg 0
Best text/layout/realism
qwen_image_2512_q8p.ckpt
recommendedPosters, typography, polished images
Creative base
z_image_1.0_q8p.ckpt
recommendedGeneral creative exploration
High-end FLUX/reference/edit
flux_2_dev_q6p.ckpt
recommendedFLUX.2 workflows on 32 GiB
Instruction image edit
qwen_image_edit_2511_q8p.ckpt
recommendedImg2img/edit prompts
Art/prompt-following alternative
hidream_i1_full_q5p.ckpt
recommended downloadHighest-quality art alternative
Fast local media
ltx_2.3_22b_distilled_q6p.ckpt
recommended downloadShort text/video generation
Rig-fit Wan media
wan_v2.2_5b_ti2v_q8p.ckpt
recommended downloadCurrent Wan video on 32 GiB
Legacy/redundant downloaded models are not deletion targets without approval:
flux_1_schnell_q8p.ckpt
,
flux_1_fill_dev_q8p.ckpt
, and
flux_2_klein_4b_q6p.ckpt
.
Auxiliary text/VLM files and incomplete partials are not recommended generation models. For example,
llama_3.1_8b_instruct_q8p.ckpt.part
is an incomplete stale auxiliary artifact, not a mid-May 2026 best-pack model.
Qwen Layered remains a current optional workflow target, but not a guaranteed best-pack install on this rig:
qwen_image_layered_1.0_bf16_q6p.ckpt
failed verification with a checksum mismatch, and
qwen_image_layered_1.0_bf16_q8p.ckpt
repeatedly stalled during CLI download. Use
qwen_image_edit_2511_q8p.ckpt
for reliable local editing unless the user explicitly asks to keep retrying Layered.
For exact install commands, current/legacy distinctions, and quantization rules, load
references/model-catalog.md
.

截至2026-05-15,该设备为Apple M1 Pro(arm64架构),运行macOS 26.5,配备32GiB统一内存。使用当前Draw Things优化后的模型 checkpoint,除非用户明确要求兼容,否则不推荐使用旧版SDXL/SD1.5/FLUX.1系列模型。
角色模型ID状态默认使用场景
快速图像生成默认模型
z_image_turbo_1.0_q8p.ckpt
推荐使用快速草稿生成,8步迭代,
--cfg 0
文本/布局/写实最佳模型
qwen_image_2512_q8p.ckpt
推荐使用海报、排版、精细化图像生成
创意基础模型
z_image_1.0_q8p.ckpt
推荐使用通用创意探索
高端FLUX参考/编辑模型
flux_2_dev_q6p.ckpt
推荐使用32GiB设备上的FLUX.2工作流
指令式图像编辑模型
qwen_image_edit_2511_q8p.ckpt
推荐使用Img2img/编辑类提示词
艺术/提示词跟随替代模型
hidream_i1_full_q5p.ckpt
推荐下载高质量艺术风格替代模型
快速本地媒体生成模型
ltx_2.3_22b_distilled_q6p.ckpt
推荐下载短文本/视频生成
适配设备的Wan媒体模型
wan_v2.2_5b_ti2v_q8p.ckpt
推荐下载32GiB设备上的当前Wan视频生成
未经用户批准,以下旧版/冗余下载模型不可删除:
flux_1_schnell_q8p.ckpt
flux_1_fill_dev_q8p.ckpt
flux_2_klein_4b_q6p.ckpt
辅助文本/VLM文件和未完成的部分文件不推荐作为生成模型。例如,
llama_3.1_8b_instruct_q8p.ckpt.part
是未完成的过时辅助文件,不属于2026年5月中旬的推荐模型集合。
Qwen Layered仍是当前可选的工作流目标,但该设备不保证默认安装推荐集合中的该模型:
qwen_image_layered_1.0_bf16_q6p.ckpt
验证时出现校验和不匹配,
qwen_image_layered_1.0_bf16_q8p.ckpt
在CLI下载时反复停滞。除非用户明确要求重试安装Layered模型,否则使用
qwen_image_edit_2511_q8p.ckpt
进行可靠的本地编辑。
如需准确的安装命令、当前/旧版模型区分规则及量化规则,请查看
references/model-catalog.md

Core Generation Protocols

核心生成流程

Every mode follows this pattern:
  1. Validate: CLI exists, input files exist, model is downloaded or
    --download-missing
    is intentional.
  2. Select defaults: choose a current model by role unless the user provided
    --model
    .
  3. Build command: use only flags shown by
    draw-things-cli generate --help
    , or use
    --config-file
    for advanced app settings.
  4. Show command: display the full command before running.
  5. Execute: run through Bash and capture output.
  6. Report: output path, seed, model id, dimensions, steps, and cfg.
所有模式均遵循以下流程:
  1. 验证:确认CLI存在、输入文件存在、模型已下载或用户明确使用
    --download-missing
    参数。
  2. 选择默认值:除非用户指定
    --model
    参数,否则根据角色选择当前推荐模型。
  3. 构建命令:仅使用
    draw-things-cli generate --help
    中显示的参数,或使用
    --config-file
    进行高级应用设置。
  4. 显示命令:执行前展示完整命令。
  5. 执行:通过Bash运行命令并捕获输出。
  6. 报告:输出路径、seed、模型ID、尺寸、迭代步数和cfg参数值。

Mode: Generate

生成模式

Fast draft default:
bash
draw-things-cli generate \
  --model z_image_turbo_1.0_q8p.ckpt \
  --prompt "<prompt>" \
  --steps 8 \
  --cfg 0 \
  --width 1024 \
  --height 1024 \
  --seed <seed> \
  --output <output.png>
Quality/text default:
bash
draw-things-cli generate \
  --model qwen_image_2512_q8p.ckpt \
  --prompt "<prompt>" \
  --steps 40 \
  --cfg 4 \
  --width 1328 \
  --height 1328 \
  --seed <seed> \
  --output <output.png>
快速草稿默认命令:
bash
draw-things-cli generate \
  --model z_image_turbo_1.0_q8p.ckpt \
  --prompt "<prompt>" \
  --steps 8 \
  --cfg 0 \
  --width 1024 \
  --height 1024 \
  --seed <seed> \
  --output <output.png>
高质量/文本优化默认命令:
bash
draw-things-cli generate \
  --model qwen_image_2512_q8p.ckpt \
  --prompt "<prompt>" \
  --steps 40 \
  --cfg 4 \
  --width 1328 \
  --height 1328 \
  --seed <seed> \
  --output <output.png>

Mode: Edit

编辑模式

bash
draw-things-cli generate \
  --model qwen_image_edit_2511_q8p.ckpt \
  --image <input.png> \
  --prompt "<edit instruction>" \
  --strength 0.55 \
  --steps 30 \
  --cfg 4 \
  --seed <seed> \
  --output <output.png>
Use
--strength 0.25-0.4
for subtle edits,
0.5-0.65
for visible edits, and
0.75+
for heavy redraws.
bash
draw-things-cli generate \
  --model qwen_image_edit_2511_q8p.ckpt \
  --image <input.png> \
  --prompt "<edit instruction>" \
  --strength 0.55 \
  --steps 30 \
  --cfg 4 \
  --seed <seed> \
  --output <output.png>
细微编辑使用
--strength 0.25-0.4
,可见编辑使用
0.5-0.65
,大幅重绘使用
0.75+

Mode: Media

媒体生成模式

bash
draw-things-cli generate \
  --model ltx_2.3_22b_distilled_q6p.ckpt \
  --prompt "<motion prompt>" \
  --frames 49 \
  --width 768 \
  --height 512 \
  --seed <seed> \
  --output <output.mov>
Use LTX distilled first for this rig. Keep width/height divisible by 32 and frames divisible by 8 plus 1.
bash
draw-things-cli generate \
  --model ltx_2.3_22b_distilled_q6p.ckpt \
  --prompt "<motion prompt>" \
  --frames 49 \
  --width 768 \
  --height 512 \
  --seed <seed> \
  --output <output.mov>
该设备优先使用LTX蒸馏模型。保持宽高为32的倍数,帧数为8的倍数加1。

Mode: Setup

设置模式

Run inventory first:
bash
uv run python skills/draw-thing/scripts/model_inventory.py --recommended-pack all --format json
Then run missing
ensure_command
entries after showing them to the user. Do not delete legacy models unless the user separately approves a prune operation.
先运行清单管理:
bash
uv run python skills/draw-thing/scripts/model_inventory.py --recommended-pack all --format json
然后向用户展示缺失模型的
ensure_command
命令并执行。除非用户单独批准清理操作,否则不得删除旧版模型。

Mode: Inventory

清单管理模式

bash
uv run python skills/draw-thing/scripts/model_inventory.py --format json
Summarize installed, missing recommended models, and redundant legacy downloads. Report
incomplete_downloads
and
auxiliary_or_dependency_files
separately so helper files or stale
.part
files are not mistaken for current best models.
bash
uv run python skills/draw-thing/scripts/model_inventory.py --format json
汇总已安装模型、缺失的推荐模型及冗余的旧版下载模型。单独报告
incomplete_downloads
auxiliary_or_dependency_files
,避免将辅助文件或过时的
.part
文件误认为当前最佳模型。

Mode: Batch

批量生成模式

The current CLI does not expose
--batch-count
. Use a shell loop and explicit output names:
bash
for seed in 1201 1202 1203 1204; do
  draw-things-cli generate \
    --model z_image_turbo_1.0_q8p.ckpt \
    --prompt "<prompt>" \
    --steps 8 \
    --cfg 0 \
    --width 1024 \
    --height 1024 \
    --seed "$seed" \
    --output "$HOME/Pictures/draw-thing/run-$seed.png"
done
当前CLI未提供
--batch-count
参数。使用Shell循环和明确的输出文件名:
bash
for seed in 1201 1202 1203 1204; do
  draw-things-cli generate \
    --model z_image_turbo_1.0_q8p.ckpt \
    --prompt "<prompt>" \
    --steps 8 \
    --cfg 0 \
    --width 1024 \
    --height 1024 \
    --seed "$seed" \
    --output "$HOME/Pictures/draw-thing/run-$seed.png"
done

Mode: Advanced

高级操作模式

The current
generate --help
does not expose old direct flags such as
--controls
,
--loras
,
--mask
,
--upscaler
,
--batch-count
,
--sampler
, or
--guidance-scale
. For ControlNet, LoRA, inpaint, upscale, tiling, and other app-level settings:
  1. Re-check
    draw-things-cli generate --help
    .
  2. Use Draw Things app-managed settings or
    --config-json
    /
    --config-file
    only when the schema is verified.
  3. Never invent direct flags from older examples.

当前
generate --help
未显示旧版直接参数,如
--controls
--loras
--mask
--upscaler
--batch-count
--sampler
--guidance-scale
。对于ControlNet、LoRA、inpaint、upscale、tiling及其他应用级设置:
  1. 重新检查
    draw-things-cli generate --help
  2. 仅在验证 schema 后,使用Draw Things应用管理的设置或
    --config-json
    /
    --config-file
    参数。
  3. 切勿从旧示例中使用未被支持的直接参数。

Prompt Quick-Reference

提示词速查表

Model familyStyleNegative prompt
Z Image Turbo/BaseDirect natural-language instructions; concise composition detailsSupported when useful, keep short
Qwen Image 2512Explicit layout, typography, exact text, editorial languageSupported, keep targeted
Qwen Edit/LayeredImperative edit instructions and preservation constraintsUse only for what to avoid changing
FLUX.2Natural language, reference/edit details, precise scene intentVerify per model; do not assume FLUX.1 rules
HiDream I1 FullRich art direction, style, medium, lighting, compositionShort targeted negatives if used
LTX/Wan mediaDescribe motion over time, camera movement, temporal continuityAvoid still-image-only tag lists
For templates and troubleshooting, load
references/prompt-patterns.md
.

模型系列风格特点负面提示词建议
Z Image Turbo/Base直接自然语言指令;简洁的构图细节必要时使用,保持简短
Qwen Image 2512明确的布局、排版、精确文本、编辑类语言支持使用,保持针对性
Qwen Edit/Layered命令式编辑指令和保留约束条件仅用于指定需避免修改的内容
FLUX.2自然语言、参考/编辑细节、精确场景意图按模型验证;勿假设遵循FLUX.1规则
HiDream I1 Full丰富的艺术指导、风格、媒介、光线、构图若使用则保持简短且针对性强
LTX/Wan媒体模型描述随时间变化的动作、相机运动、时间连续性避免仅使用静态图像标签列表
如需模板和故障排查,请查看
references/prompt-patterns.md

Scaling Strategy

适配策略

Request sizeStrategy
Single imageUse
z_image_turbo_1.0_q8p.ckpt
unless quality/text needs Qwen 2512
Text/layout criticalUse
qwen_image_2512_q8p.ckpt
and explicit text/layout prompt
Edit existing imageUse
qwen_image_edit_2511_q8p.ckpt
with conservative strength first
Layer/editability workflowUse Qwen Layered only if already installed; otherwise explain prior install failures and ask before retrying
Heavy art qualityTry
hidream_i1_full_q5p.ckpt
; fall back if memory pressure appears
Short videoUse LTX distilled q6p first; keep dimensions moderate
Many variationsUse shell loops, fixed output paths, and modest seed batches

请求类型策略说明
单张图像生成默认使用
z_image_turbo_1.0_q8p.ckpt
,若对质量/文本要求高则使用Qwen 2512
文本/布局要求严格使用
qwen_image_2512_q8p.ckpt
并搭配明确的文本/布局提示词
编辑现有图像先使用
qwen_image_edit_2511_q8p.ckpt
并设置保守的strength值
分层/可编辑工作流仅在已安装Qwen Layered时使用;否则说明之前的安装失败并询问用户是否重试
高质量艺术风格生成尝试使用
hidream_i1_full_q5p.ckpt
;若出现内存压力则切换其他模型
短视频生成优先使用LTX蒸馏q6p模型;保持尺寸适中
多版本变体生成使用Shell循环、固定输出路径和适量的seed批次

Output Handling

输出处理

  • Default skill output directory:
    ~/Pictures/draw-thing/
    .
  • Create it before generation:
    mkdir -p "$HOME/Pictures/draw-thing"
    .
  • Always pass
    --output <file>
    ; do not rely on implicit output behavior.
  • Never overwrite source images; use descriptive new filenames.
  • Always report output file, seed, model, steps, cfg, dimensions, and whether the model was downloaded during the run.

  • 默认技能输出目录:
    ~/Pictures/draw-thing/
  • 生成前创建目录:
    mkdir -p "$HOME/Pictures/draw-thing"
  • 必须传入
    --output <file>
    参数;不要依赖隐式输出行为。
  • 切勿覆盖源图像;使用描述性的新文件名。
  • 必须报告输出文件、seed、模型、迭代步数、cfg参数、尺寸以及模型是否在本次运行中下载。

Error Recovery

错误恢复

ErrorLikely causeAction
unknown flag
Old guidance or CLI mismatchRun
draw-things-cli generate --help
; remove stale direct flag
model not foundWrong id or not downloadedRun
draw-things-cli models list
; use
models ensure --model <id>
no output fileMissing
--output
or preview-only terminal mode
Re-run with explicit file path
killed/OOMModel too large or dimensions too highLower dimensions, use q6p/q5p, or switch to faster model
poor textWrong model/prompt for typographyUse Qwen Image 2512 with exact quoted text
video failsModel missing or memory pressureEnsure LTX q6p, reduce dimensions/frames

错误类型可能原因处理措施
unknown flag
旧版引导参数或CLI版本不匹配运行
draw-things-cli generate --help
;移除过时的直接参数
model not found模型ID错误或未下载运行
draw-things-cli models list
;使用
models ensure --model <id>
命令
no output file缺失
--output
参数或终端仅预览
使用明确的文件路径重新运行
killed/OOM模型过大或尺寸太高减小尺寸、使用q6p/q5p模型或切换更快的模型
poor text模型/提示词不适合排版生成使用Qwen Image 2512模型并搭配带引号的精确文本提示词
video fails模型缺失或内存压力过大确保安装LTX q6p模型、减小尺寸/帧数

Reference Files

参考文件

Load ONE reference at a time.
FileContentLoad When
references/cli-reference.md
Current command surface, flags, models subcommands, config caveatsBuilding non-trivial commands
references/model-catalog.md
2026-05-15 current best model pack, exact ensure commands, legacy notesChoosing/downloading models
references/prompt-patterns.md
Prompt patterns for Qwen, Z Image, FLUX.2, HiDream, LTX/WanComplex prompts or quality issues
references/controlnet-guide.md
Advanced controls/import/config guidance with stale-flag warningsControlNet, LoRA, inpaint, upscale
references/workflow-recipes.md
Current recipes using exposed flags and shell loopsMulti-step creative workflows

每次仅加载一个参考文件。
文件路径内容说明加载时机
references/cli-reference.md
当前命令集、参数、models子命令、配置注意事项构建非简单命令时
references/model-catalog.md
2026-05-15当前最佳模型集合、准确的ensure命令、旧版模型说明选择/下载模型时
references/prompt-patterns.md
Qwen、Z Image、FLUX.2、HiDream、LTX/Wan的提示词模式处理复杂提示词或质量问题时
references/controlnet-guide.md
高级控制/导入/配置指南及过时参数警告处理ControlNet、LoRA、inpaint、upscale时
references/workflow-recipes.md
使用已公开参数和Shell循环的当前工作流示例处理多步骤创意工作流时

Validation Contract

验证规范

Run from this skill directory before declaring changes complete:
bash
python scripts/check.py
python scripts/model_inventory.py --format json
Completion criteria:
  1. scripts/check.py
    exits 0.
  2. Model inventory identifies recommended-pack status.
  3. No command examples use stale direct flags unsupported by current
    generate --help
    .
  4. No portable-CLI violations remain under this skill directory.

在完成变更前,从技能目录运行以下命令:
bash
python scripts/check.py
python scripts/model_inventory.py --format json
完成标准:
  1. scripts/check.py
    执行返回0。
  2. 模型清单能识别推荐模型集合的状态。
  3. 所有命令示例均未使用当前
    generate --help
    不支持的过时直接参数。
  4. 技能目录下无违反可移植CLI规范的内容。

Critical Rules

关键规则

  1. Always check
    command -v draw-things-cli
    before generation or model mutation.
  2. Always use
    draw-things-cli generate --help
    as the source of truth for direct flags.
  3. Always pass
    --output
    ; preview-only output is not enough for agent workflows.
  4. Use
    --cfg
    , not stale
    --guidance-scale
    .
  5. Do not use stale direct flags:
    --sampler
    ,
    --batch-count
    ,
    --upscaler
    ,
    --mask
    ,
    --controls
    , or
    --loras
    unless current help/config schema proves support.
  6. Prefer the recommended 2026 rig pack over legacy SDXL, SD1.5, SD2.x, SVD, FLUX.1, Qwen 1.0, Qwen Edit 2509, Wan 2.1, or old community checkpoints.
  7. Never delete or prune models without separate explicit user approval.
  8. Show the full command before running it.
  9. Preserve source images; write new output files.
  10. Report seed, model id, output path, dimensions, steps, and cfg.
  11. For video, keep dimensions moderate and frame count divisible by 8 plus 1.
  12. For advanced ControlNet/LoRA/inpaint/upscale workflows, verify
    --config-json
    or
    --config-file
    schema before execution.
  13. Do not treat auxiliary text/VLM files, sidecars, or incomplete
    .part
    downloads as recommended generation/media models.
  1. 在执行生成或模型变更操作前,必须检查
    command -v draw-things-cli
  2. 必须以
    draw-things-cli generate --help
    作为直接参数的唯一来源。
  3. 必须传入
    --output
    参数;仅预览输出不足以满足Agent工作流需求。
  4. 使用
    --cfg
    参数,而非过时的
    --guidance-scale
  5. 不得使用过时的直接参数:
    --sampler
    --batch-count
    --upscaler
    --mask
    --controls
    --loras
    ,除非当前帮助/配置schema确认支持。
  6. 优先使用2026年设备推荐模型集合,而非旧版SDXL、SD1.5、SD2.x、SVD、FLUX.1、Qwen 1.0、Qwen Edit 2509、Wan 2.1或旧社区checkpoint。
  7. 未经用户单独明确批准,不得删除或清理模型。
  8. 执行前展示完整命令。
  9. 保留源图像;生成新的输出文件。
  10. 报告seed、模型ID、输出路径、尺寸、迭代步数和cfg参数。
  11. 生成视频时,保持尺寸适中且帧数为8的倍数加1。
  12. 对于高级ControlNet/LoRA/inpaint/upscale工作流,执行前需验证
    --config-json
    --config-file
    的schema。
  13. 不得将辅助文本/VLM文件、附属文件或未完成的
    .part
    下载文件视为推荐的生成/媒体模型。