LiblibAI Image & Video Gen

Generate images with Seedream4.5 and videos with Kling via LiblibAI API. Use when user asks to generate/create images, pictures, illustrations, or videos using LiblibAI, Seedream, or Kling models.

Security Scan

The skill's code and instructions match its stated image/video generation purpose, but the package metadata omits required credentials (LIB_ACCESS_KEY / LIB_SECRET_KEY) and the skill comes from an unknown source — this mismatch is a red flag worth investigating before installing.

Purpose & Capability

The SKILL.md and bundled Python client clearly implement LiblibAI image (Seedream4.5) and video (Kling) generation and require API credentials — that aligns with the stated purpose. However, the registry metadata declares no required environment variables or primary credential while the runtime instructions and code require LIB_ACCESS_KEY and LIB_SECRET_KEY. The omission in metadata is an incoherence that reduces transparency.

Instruction Scope

SKILL.md instructs the agent to run the included Python CLI which authenticates and posts generation requests to https://openapi.liblibai.cloud and polls job status. The instructions and code only reference the declared API endpoints, prompt parameters, and the two API env vars; they do not attempt to read unrelated files, other env vars, or send data to unexpected endpoints.

Install Mechanism

There is no install specification (instruction-only with a bundled script). That is low-risk compared with arbitrary downloads. The package includes a single Python script relying on requests; no external installers or obscure URLs are used.

Credentials

The code legitimately needs an API access key and secret (LIB_ACCESS_KEY and LIB_SECRET_KEY) to call the service — that is proportionate to the function. However, the skill registry metadata lists no required env vars; this discrepancy could trick automated gates or users into installing without realizing they must supply secrets. The only sensitive data the skill uses is the two API credentials (no other secrets appear in code).

Persistence & Privilege

The skill does not request always:true, does not modify other skills or system config, and runs only when invoked. It does not request elevated persistence or system-wide privileges.

What to consider before installing

Before installing, verify the skill's origin (no homepage/source provided) and confirm you trust LiblibAI and the owner. The skill requires two sensitive env vars (LIB_ACCESS_KEY and LIB_SECRET_KEY) that will be sent to https://openapi.liblibai.cloud — only provide keys you control and are comfortable using with a third-party skill. Ask the publisher to update registry metadata to declare the required env vars (transparency improves security). If you cannot verify the source, prefer an officially published client or use short-lived credentials and monitor their use. If you install, inspect or run the included script in a sandbox first (it uses HMAC-SHA1 signing and posts job requests and polling calls; there is no other hidden exfiltration in the code).

LiblibAI Image & Video Generation

Generate images (Seedream4.5) and videos (Kling) via LiblibAI's API.

Prerequisites

Environment variables must be set:

• LIB_ACCESS_KEY — API access key
• LIB_SECRET_KEY — API secret key

Usage

Run the CLI script at scripts/liblib_client.py:

# Generate image
python3 scripts/liblib_client.py image "a cute cat wearing a hat" --width 2048 --height 2048

# Generate video from text
python3 scripts/liblib_client.py text2video "a rocket launching into space" --model kling-v2-6 --duration 5

# Generate video from image
python3 scripts/liblib_client.py img2video "the cat turns its head" --start-frame https://example.com/cat.jpg

# Check task status
python3 scripts/liblib_client.py status <generateUuid>

Image Generation (Seedream4.5)

• Endpoint: POST /api/generate/seedreamV4
• Model: doubao-seedream-4-5-251128
• Default size: 2048×2048. For 4.5, min total pixels = 3,686,400 (e.g. 2560×1440)
• Supports reference images (1-14), prompt expansion, and sequential image generation
• Options: --width, --height, --count (1-15), --ref-images, --prompt-magic

Video Generation (Kling)

Text-to-Video

• Endpoint: POST /api/generate/video/kling/text2video
• Models: kling-v2-6 (latest, supports sound), kling-v2-1-master, kling-v2-5-turbo, etc.
• Options: --model, --aspect (16:9/9:16/1:1), --duration (5/10s), --mode (std/pro), --sound (on/off)

Image-to-Video

• Endpoint: POST /api/generate/video/kling/img2video
• Provide --start-frame image URL; optionally --end-frame (v1-6 only)
• For kling-v2-6: uses images array instead of startFrame/endFrame

Async Pattern

All generation tasks are async:

1. Submit task → get generateUuid
2. Poll POST /api/generate/status with { "generateUuid": "..." }
3. Result contains images[].imageUrl or videos[].videoUrl

The script auto-polls by default. Use --no-poll to submit only.

Notes

• QPS limit: 1 request/second for task submission
• Max concurrent tasks: 5
• Image URLs in results expire after 7 days
• For kling-v2-5-turbo and kling-v2-6, mode must be "pro" (default)
• Sound generation only supported on kling-v2-6+