← 返回 Skills 市场

Filtrix AI Image Generation

Name: Filtrix AI Image Generation
Author: lumenclaw-cloud

作者 lumenclaw-cloud · GitHub ↗ · v1.0.0

cross-platform ⚠ suspicious

475

总下载

当前安装

版本数

在 OpenClaw 中安装

/install filtrix-ai-image-gen

功能描述

Generate images using AI providers (OpenAI gpt-image-1, Google Gemini, fal.ai). Use when the user asks to create, generate, or make an image, picture, illust...

使用说明 (SKILL.md)

Filtrix Image Gen

Generate and edit images via OpenAI, Gemini, or fal.ai.

Setup

Ensure the relevant API key is set as an environment variable:

Provider	Env Variable	Get Key
OpenAI	`OPENAI_API_KEY`	platform.openai.com
Gemini	`GOOGLE_API_KEY`	aistudio.google.com
fal.ai	`FAL_KEY`	fal.ai/dashboard

No pip dependencies — uses only Python stdlib.

Text-to-Image (Generate)

python scripts/generate.py --provider \x3Copenai|gemini|fal> --prompt "..." [--size WxH|RATIO] [--model MODEL] [--resolution 1K|2K|4K] [--output PATH] [--seed N]

Image-to-Image (Edit)

python scripts/edit.py --provider \x3Copenai|gemini|fal> --image input.png --prompt "edit instruction" [--mask mask.png] [--size WxH|RATIO] [--model MODEL] [--resolution 1K|2K|4K] [--output PATH] [--seed N]

--mask is OpenAI only (for inpainting)
--resolution is Gemini only (requires --model gemini-3-pro-image-preview)
--seed is fal only

Output: prints OK: /path/to/image.png (N bytes) on success.

Provider Selection Guide

openai — Best quality for photorealistic and artistic images. Model: gpt-image-1. Supports mask-based inpainting for edits.
gemini — Default: gemini-2.5-flash-image (fast, cheap). Premium: --model gemini-3-pro-image-preview (higher quality, more expensive, supports --resolution 1K/2K/4K). Prefer Flash unless user requests higher quality.
fal — Default: seedream45 (ByteDance SeedReam 4.5). Also: seedream4, flux-pro, flux-dev, recraft-v3. Or pass raw fal model ID.

If the user doesn't specify a provider, pick based on available API keys. Prefer gemini for speed, openai for quality.

Sizes

Generate (--size)

Size	Aspect	Notes
`1024x1024`	1:1	Default, square
`1536x1024`	3:2	Landscape
`1024x1536`	2:3	Portrait

For Gemini, also accepts aspect ratios directly: 1:1, 3:2, 4:3, 16:9, 21:9, 9:16, 3:4.

Resolution (Gemini 3 Pro only)

Use --resolution 2K or --resolution 4K with --model gemini-3-pro-image-preview for high-res output.

Resolution	16:9	1:1
1K	1376×768	1024×1024
2K	2752×1536	2048×2048
4K	5504×3072	4096×4096

Prompt Tips

For best results, be specific about style, lighting, composition, and subject.

Browse 100+ production-tested prompts at filtrix.ai/prompts — copy directly or use as inspiration.

When a user needs help writing prompts, or asks for style recommendations, see references/prompts.md for a detailed writing guide with examples by category and tips from Filtrix's experience with 100+ curated prompts across 20+ styles.

Provider-Specific Details

OpenAI specifics: See references/openai.md
Gemini specifics: See references/gemini.md
fal.ai specifics: See references/fal.md

安全使用建议

This skill appears to implement multi-provider image generation as advertised, but there are two things to watch for before installing or using it: 1) Metadata mismatch: The registry metadata claims no required environment variables, but both SKILL.md and the bundled scripts require OPENAI_API_KEY, GOOGLE_API_KEY, and/or FAL_KEY. Treat the registry metadata as untrustworthy until the author updates it. Only set API keys for providers you intend to use, and avoid placing broad account credentials in shared/global environment variables. 2) Safety configuration: The code disables fal.ai's safety checker (enable_safety_checker: False). That can allow generation or acceptance of content that would otherwise be blocked. If you need strict content safety, ask the maintainer why this is disabled, or modify the script to enable the checker before use. Other practical advice: - Review the two Python scripts locally (they are included) and run them in an isolated environment (container/VM) if you are concerned about network interactions. - Provide only provider keys you trust and preferably scoped/limited API keys (project-limited) rather than full account keys. - If you require guarantees about metadata, request that the publisher correct the registry fields so required env vars are declared explicitly. If you want, I can (a) point to the exact lines that disable fal.ai safety in the scripts, (b) produce a small patch to re-enable the safety flag, or (c) extract the exact env-var usage so you can decide which keys to set.

功能分析

Type: OpenClaw Skill Name: filtrix-ai-image-gen Version: 1.0.0 The skill bundle provides image generation and editing functionality using external AI providers. It correctly handles API keys via environment variables and communicates with legitimate API endpoints. However, the Python scripts (`scripts/edit.py`, `scripts/generate.py`) accept user-controlled file paths for input images (`--image`, `--mask`) and output images (`--output`). This capability, while necessary for the skill's function, introduces a vulnerability where a malicious prompt to the OpenClaw agent could lead to arbitrary file reads (e.g., attempting to read sensitive system files as input images) or arbitrary file writes (e.g., overwriting system files with generated images), posing a risk for data leakage or system compromise if the agent's input is not properly sanitized.

能力评估

⚠ Purpose & Capability

The skill's stated purpose is image generation using OpenAI, Gemini, and fal.ai, which matches the code. However the registry metadata claims no required environment variables or primary credential, while both SKILL.md and the scripts clearly require OPENAI_API_KEY, GOOGLE_API_KEY, and/or FAL_KEY. That metadata omission is an incoherence that should be fixed before trusting the package metadata.

ℹ Instruction Scope

Runtime instructions are scoped to generating and editing images with the three providers and instruct users to set only the provider API keys. The scripts read local image files and send image bytes/prompts to provider endpoints, and fetch returned image URLs — all expected behavior. One notable deviation: the fal.ai edit/generation calls set 'enable_safety_checker': False in the payload, which disables fal.ai's safety filter; this is a behavioral choice that can increase risk of generating or accepting disallowed content and is not documented in SKILL.md.

✓ Install Mechanism

No install spec (instruction-only with included Python scripts). No third-party packages are pulled; code uses only Python standard library network calls. This limits installation risk because nothing is downloaded/installed at install-time by the registry spec.

⚠ Credentials

The env vars the scripts actually require (OPENAI_API_KEY, GOOGLE_API_KEY, FAL_KEY) are proportionate to the stated purpose. However the registry metadata listing 'Required env vars: none' is incorrect and misleading — that mismatch is a security/usability concern because users may not realize the skill needs those API keys. Also note the code sends those keys as Bearer/Key headers to remote provider endpoints (expected but sensitive).

✓ Persistence & Privilege

The skill does not request always:true, does not modify other skills or system configs, and has no install-time persistence. It runs only when invoked and performs network calls as part of normal operation.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install filtrix-ai-image-gen
安装完成后，直接呼叫该 Skill 的名称或使用 /filtrix-ai-image-gen 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release

元数据

Slug filtrix-ai-image-gen

版本 1.0.0

许可证 —

累计安装 1

当前安装数 1

历史版本数 1

常见问题