// Generate/edit images with OpenAI gpt-image-2 by default, falling back to Gemini (gemini-3.1-flash-image-preview) when OPENAI_API_KEY is unset. Supports text-to-image + image-to-image; 1K/2K/4K; use --input-image for editing, --provider to force a provider, --model to override the model.
| name | inno-figure-gen |
| description | Generate/edit images with OpenAI gpt-image-2 by default, falling back to Gemini (gemini-3.1-flash-image-preview) when OPENAI_API_KEY is unset. Supports text-to-image + image-to-image; 1K/2K/4K; use --input-image for editing, --provider to force a provider, --model to override the model. |
Generate new images or edit existing ones. The script picks a provider based on which API keys are available:
gpt-image-2 — used when OPENAI_API_KEY is set (default).gemini-3.1-flash-image-preview — used when only GEMINI_API_KEY is set.When both keys are set, OpenAI is picked by default; force Gemini with --provider gemini. If an auto-selected OpenAI call fails at runtime (quota, moderation, network), the script transparently falls back to Gemini when a Gemini key is available.
The script is at scripts/generate_image.py relative to this skill's directory (the directory containing this SKILL.md). Resolve the full path from the skill's location before running. Do not hardcode ~/.codex/..., because the skill may be installed in a different location.
Keep the distinction clear:
generate_image.py.--filename.Generate new image:
uv run <this-skill-directory>/scripts/generate_image.py --prompt "your image description" --filename "output-name.png" [--resolution 1K|2K|4K] [--provider auto|openai|gemini] [--model MODEL] [--openai-api-key KEY | --gemini-api-key KEY]
Edit existing image:
uv run <this-skill-directory>/scripts/generate_image.py --prompt "editing instructions" --filename "output-name.png" --input-image "path/to/input.png" [--resolution 1K|2K|4K] [--provider auto|openai|gemini] [--model MODEL] [--openai-api-key KEY | --gemini-api-key KEY]
Important: Always run from the user's current working directory so images are saved where the user is working, not in the skill directory.
Goal: fast iteration without burning time on 4K until the prompt is correct.
uv run <this-skill-directory>/scripts/generate_image.py --prompt "<draft prompt>" --filename "yyyy-mm-dd-hh-mm-ss-draft.png" --resolution 1K--input-image for every iteration until you’re happy.uv run <this-skill-directory>/scripts/generate_image.py --prompt "<final prompt>" --filename "yyyy-mm-dd-hh-mm-ss-final.png" --resolution 4KThe script accepts three resolution tiers (uppercase K required):
Map user requests to API parameters:
1K1K2K4KOpenAI 4K note: gpt-image-2 supports non-square 4K outputs within its size limits. --resolution 4K with OpenAI maps to 3840×2160.
Two providers are available:
| Provider | Default model | When chosen |
|---|---|---|
| OpenAI | gpt-image-2 | --provider auto (default) when OPENAI_API_KEY is set, or --provider openai |
| Gemini | gemini-3.1-flash-image-preview | --provider auto when only GEMINI_API_KEY is set, or --provider gemini, or as runtime fallback from a failed auto-OpenAI call |
Override either default with --model. Note: the model name is provider-specific; passing a Gemini model name while the script falls back to Gemini automatically will not preserve a user-specified OpenAI model (each provider uses its own default during fallback).
Common model options:
gpt-image-2 (default)gemini-3.1-flash-image-preview (default, fast), gemini-3-pro-image-preview (higher quality, slower)The script resolves provider-specific keys first, while preserving the original Gemini-only --api-key behavior:
--openai-api-key KEY, --gemini-api-key KEY--api-key KEY — kept for backward compatibility with the original Gemini-only script. Under auto, it is treated as a Gemini key unless an OpenAI key is provided by --openai-api-key or OPENAI_API_KEY. Under explicit --provider openai, it is treated as an OpenAI key; under explicit --provider gemini, it is treated as a Gemini key.OPENAI_API_KEY, GEMINI_API_KEYWhen --provider auto is selected and OpenAI fails at runtime, fallback to Gemini requires --gemini-api-key, --api-key, or the GEMINI_API_KEY env var.
If no key is resolvable for the chosen provider, the script exits with a clear error message listing both ways to fix it.
Preflight:
command -v uv (must exist)test -n "$OPENAI_API_KEY" -o -n "$GEMINI_API_KEY" (or pass --openai-api-key, --gemini-api-key, or backward-compatible --api-key)test -f "path/to/input.png"Common failures:
Error: No API key found... → set OPENAI_API_KEY or GEMINI_API_KEY, or pass an explicit --*-api-key flagError loading input image: → wrong path / unreadable file; verify --input-image points to a real image[warn] OpenAI call failed (...); falling back to Gemini. → informational; a Gemini key was available and produced the image. Investigate the OpenAI error separately (quota, moderation, network)--provider openai (explicit provider disables fallback). Try a different key, or drop the explicit provider to enable fallbackGenerate filenames with the pattern: yyyy-mm-dd-hh-mm-ss-name.png
Format: {timestamp}-{descriptive-name}.png
yyyy-mm-dd-hh-mm-ss (24-hour format)x9k2, a7b3)Examples:
2025-11-23-14-23-05-japanese-garden.png2025-11-23-15-30-12-sunset-mountains.png2025-11-23-16-45-33-robot.png2025-11-23-17-12-48-x9k2.pngWhen the user wants to modify an existing image:
--input-image parameter with the path to the imageFor generation: Pass user's image description as-is to --prompt. Only rework if clearly insufficient.
For editing: Pass editing instructions in --prompt (e.g., "add a rainbow in the sky", "make it look like a watercolor painting")
Preserve user's creative intent in both cases.
Use templates when the user is vague or when edits must be precise.
Generation template:
Editing template (preserve everything else):
Generate new image (auto provider):
uv run <this-skill-directory>/scripts/generate_image.py --prompt "A serene Japanese garden with cherry blossoms" --filename "2025-11-23-14-23-05-japanese-garden.png" --resolution 2K
Force Gemini:
uv run <this-skill-directory>/scripts/generate_image.py --prompt "A serene Japanese garden with cherry blossoms" --filename "2025-11-23-14-23-05-japanese-garden-4k.png" --resolution 4K --provider gemini
Edit existing image (auto provider):
uv run <this-skill-directory>/scripts/generate_image.py --prompt "make the sky more dramatic with storm clouds" --filename "2025-11-23-14-25-30-dramatic-sky.png" --input-image "original-photo.jpg" --resolution 2K
Mandatory pre-flight compute resource check before running experiments. Detects whether local/remote GPU or compute resources are actually available. If resources are unavailable, STOPS the experiment pipeline immediately and reports to the user — preventing the model from hallucinating fake experiment results. Use when: about to run experiments, deploy training, or any GPU-intensive task.
Workflow 1.5: Bridge between idea discovery and auto review. Reads EXPERIMENT_PLAN.md, implements experiment code, deploys to GPU, collects initial results. Use when user says "实现实验", "implement experiments", "bridge", "从计划到跑实验", "deploy the plan", or has an experiment plan ready to execute.
Deploy and run ML experiments on local, remote, Vast.ai, or Modal serverless GPU. Use when user says "run experiment", "deploy to server", "跑实验", or needs to launch training jobs.
Dr. Claw workspace skill for project lookup, session inspection, TaskMaster progress, OpenClaw structured schema, and event-driven reporting
Use when a quest needs one or more follow-up runs such as ablations, robustness checks, error analysis, or failure analysis after a main experiment.
Use when a quest needs to attach, import, reproduce, repair, verify, compare, or publish a baseline and its metrics.