OpenAI Image CLI

Generate, edit, and manage images via OpenAI's GPT Image and DALL-E models.

Security Scan

The skill's stated purpose (an OpenAI image CLI) matches its runtime instructions, but there are inconsistencies between the registry metadata and the SKILL.md (notably around required binaries/env and install behavior), and the skill relies on installing a third-party npm package without an install spec—review before installing.

Purpose & Capability

The SKILL.md describes a CLI that uses the OpenAI API and expects an OPENAI_API_KEY and an 'openai-image' binary — that is coherent with an image-generation CLI. However the registry metadata (requirements block shown earlier) lists no required env vars or binaries, while the SKILL.md metadata declares bins: ["openai-image"] and envs: ["OPENAI_API_KEY"]. This mismatch is an incoherence that could cause surprising behavior (the agent may assume no credentials are needed while the runtime expects them).

Instruction Scope

The instructions are focused on installing and using an external CLI, using local image files, stdin, and a local history/config. They do not direct the agent to exfiltrate arbitrary host files or read unrelated system state. Still, the SKILL.md instructs installation of a third-party npm CLI and expects the agent to invoke that binary — that expands runtime scope beyond an instruction-only skill and should be explicitly declared in the registry.

Install Mechanism

There is no install spec in the registry, but the SKILL.md tells users to run 'npm install -g @versatly/openai-image-cli'. Relying on a user-installed npm package is a legitimate design choice but the absence of an explicit install spec in the skill metadata is an incoherence. Installing a global npm package pulls code from the public registry (moderate risk); the package and its GitHub repo should be audited before installing.

Credentials

The only credential actually needed by the CLI is OPENAI_API_KEY, which is appropriate for this purpose. But the registry's declared required env vars is empty while SKILL.md expects OPENAI_API_KEY; this discrepancy could mislead users about credential needs. No other unrelated secrets are requested.

Persistence & Privilege

No elevated privileges are requested: always is not set, model invocation is not disabled, and there are no required config paths. The skill does not request permanent presence or extra privileges.

What to consider before installing

This skill appears to be a wrapper around a third-party npm CLI (@versatly/openai-image-cli) and legitimately needs an OPENAI_API_KEY. Before installing or using it: 1) Verify the npm package and GitHub repository (review code, recent commits, and maintainers) to ensure it does nothing unexpected. 2) Be aware the SKILL.md expects you to install a global npm package even though the registry lists no install spec—the agent will not auto-install it. 3) Use a scoped OpenAI key or billing limits if possible (don't expose a full-access key). 4) Avoid installing global packages with elevated privileges (sudo) unless you trust the source. 5) If you need stronger guarantees, request the skill author add an explicit install spec and correct the declared requirements (bins/envs) in the registry so the skill's metadata matches its runtime instructions.

Generate, edit, and create variations of images using OpenAI's latest models.

Installation

npm install -g @versatly/openai-image-cli

Authentication

# Via environment variable
export OPENAI_API_KEY=sk-...

# Or via config
openai-image config set api-key sk-...

Quick Start

# Generate an image
openai-image generate "A futuristic city at sunset"

# High quality landscape
openai-image generate "Mountain panorama" -s 1536x1024 -q high

# Multiple images with transparency
openai-image generate "Logo design" -n 4 -b transparent

# Edit an existing image
openai-image edit photo.png "Add sunglasses to the person"

# Create variations (DALL-E 2)
openai-image vary original.png -n 3

Available Models

Model	Description	Notes
gpt-image-1.5	Latest GPT Image (default)	Best quality, recommended
gpt-image-1	GPT Image	Good balance
gpt-image-1-mini	GPT Image Mini	Cost-effective
dall-e-3	DALL-E 3	Deprecated May 2026
dall-e-2	DALL-E 2	Deprecated May 2026, supports variations

Commands

generate

Create images from text prompts.

openai-image generate "prompt" [options]

Options:
  -m, --model <model>        Model (default: gpt-image-1.5)
  -s, --size <size>          Size: 1024x1024, 1536x1024, 1024x1536, auto
  -q, --quality <quality>    Quality: auto, high, medium, low
  -n, --count <n>            Number of images (1-10)
  -f, --format <format>      Format: png, jpeg, webp
  -o, --output <path>        Output file/directory
  -b, --background <bg>      Background: auto, transparent, opaque
  --compression <0-100>      Compression level for jpeg/webp
  --moderation <level>       Content moderation: auto, low
  --stream                   Enable streaming with partial images
  --partial-images <0-3>     Partial images during streaming
  --json                     Output JSON response
  --dry-run                  Show request without executing

edit

Edit existing images with prompts.

openai-image edit <image> "instructions" [options]

Options:
  --mask <path>              Mask image for inpainting
  --images <paths...>        Additional reference images (up to 16)
  -s, --size <size>          Output size
  -q, --quality <quality>    Quality level
  -n, --count <n>            Number of variations
  -f, --format <format>      Output format
  -o, --output <path>        Output path

Examples:

# Simple edit
openai-image edit photo.png "Add sunglasses"

# Inpainting with mask
openai-image edit room.png "Add a plant" --mask mask.png

# Multi-image composite
openai-image edit base.png "Create gift basket" --images item1.png item2.png

vary

Create variations of an image (DALL-E 2 only).

openai-image vary <image> [options]

Options:
  -n, --count <n>            Number of variations (1-10)
  -s, --size <size>          Size: 256x256, 512x512, 1024x1024
  -o, --output <path>        Output path/directory

batch

Generate multiple images from a file or stdin.

openai-image batch [options]

Options:
  -i, --input <file>         Input file (text or JSONL)
  --stdin                    Read from stdin
  -m, --model <model>        Model for all generations
  -o, --output-dir <dir>     Output directory
  --parallel <n>             Concurrent requests (default: 3)
  --delay <ms>               Delay between requests (default: 100)

JSONL format:

{"prompt": "A red car", "size": "1024x1024", "quality": "high"}
{"prompt": "A blue boat", "size": "1536x1024"}

config

Manage CLI configuration.

openai-image config set <key> <value>
openai-image config get <key>
openai-image config list
openai-image config reset
openai-image config path

Keys: api-key, default-model, default-size, default-quality, default-format, output-dir

models

List available models.

openai-image models [--json]

history

View local generation history.

openai-image history [-n <limit>] [--json] [--clear]

Output Formats

Default (human-readable)

✓ Generated image saved to ./generated-1707500000.png
  Model: gpt-image-1.5
  Size: 1024x1024
  Quality: high
  Tokens: 150 (text: 10, image: 140)

JSON (--json)

{
  "success": true,
  "file": "./generated-1707500000.png",
  "model": "gpt-image-1.5",
  "size": "1024x1024",
  "quality": "high",
  "usage": {
    "total_tokens": 150,
    "input_tokens": 50,
    "output_tokens": 100
  }
}

Size Options

Model	Sizes
GPT Image	1024x1024, 1536x1024 (landscape), 1024x1536 (portrait), auto
DALL-E 3	1024x1024, 1792x1024, 1024x1792
DALL-E 2	256x256, 512x512, 1024x1024

Tips

1. Transparent backgrounds: Use -b transparent -f png for logos
2. Batch processing: Use JSONL for per-image options
3. Cost control: Use gpt-image-1-mini for drafts
4. History tracking: Enabled by default, view with openai-image history

Links

• npm: https://www.npmjs.com/package/@versatly/openai-image-cli
• GitHub: https://github.com/Versatly/openai-image-cli