← Back to Skills Marketplace
Sogni Gen
by
Mauvis Ledford
· GitHub ↗
· v1.5.16
· MIT-0
955
Downloads
2
Stars
1
Active Installs
16
Versions
Install in OpenClaw
/install sogni-agent
Description
Generate images **and videos** using Sogni AI's decentralized network, with local credential/config files and optional local media inputs. Ask the agent to "...
Usage Guidance
This skill appears to be what it claims: a Sogni AI image/video generator that uses a local Node CLI and an optional MCP wrapper. Before installing: 1) Be prepared to provide your SOGNI_API_KEY (or username/password) either in ~/.config/sogni/credentials or via environment variables; 2) note that generated assets are saved to ~/Downloads/sogni by default and last-render metadata is written to ~/.config/sogni/last-render.json — you can override or disable these via the documented env vars (SOGNI_DOWNLOADS_DIR, SOGNI_MCP_SAVE_DOWNLOADS); 3) the MCP server will only auto-download result URLs from sogni.ai by default, but SOGNI_ALLOWED_DOWNLOAD_HOSTS can broaden that — avoid setting it to permit arbitrary hosts unless you understand the implications; 4) installing runs npm install for the packaged runtime deps — review package.json/skill-package.json if you want to audit dependencies. Overall, the components and privileges requested match the stated purpose.
Capability Analysis
Type: OpenClaw Skill
Name: sogni-gen
Version: 1.5.16
The sogni-gen skill bundle provides a legitimate interface for generating images and videos via Sogni AI. The implementation includes several security-conscious features, such as path sanitization to prevent shell injection, directory traversal checks in the media listing logic (sogni-gen.mjs), and hostname validation for automatic downloads (mcp-server.mjs). Furthermore, the SKILL.md instructions explicitly direct the AI agent to use provided safe wrappers instead of executing raw shell commands like ffmpeg or ls, reducing the risk of accidental or malicious system compromise.
Capability Assessment
Purpose & Capability
The name/description (Sogni image & video generation) matches the actual behavior: a Node CLI + MCP wrapper that calls a Sogni client library. Required binaries (node, optional ffmpeg) and deps in package.json logically support image/video generation. The requested env vars and config paths (SOGNI_API_KEY, username/password, credentials path, OpenClaw config, inbound media dir, downloads dir) are appropriate for a plugin that integrates with Sogni and OpenClaw.
Instruction Scope
Runtime instructions and code read Sogni credentials (file or env), read/write a last-render.json, read OpenClaw config and an inbound media directory, and may download generated result URLs and save them to a local Downloads directory. The MCP server sanitizes inputs and restricts downloads to trusted hosts by default (sogni.ai). Note: the SOGNI_ALLOWED_DOWNLOAD_HOSTS environment variable can expand which HTTPS hosts are permitted for auto-downloads — if misconfigured this could allow downloading from arbitrary hosts. Otherwise the instructions stay within the stated purpose.
Install Mechanism
The skill uses npm to install runtime dependencies (skill-package.json -> npm i). Dependencies are typical (Sogni client wrapper, MCP SDK, sharp, execa). No archive downloads from arbitrary URLs or shell-executed remote scripts are present in the manifest. (Minor metadata inconsistency: registry summary said 'No install spec', but SKILL.md includes an npm install step — this is explanatory rather than harmful.)
Credentials
Requested environment variables are primarily Sogni credentials, OpenClaw configuration hooks, ffmpeg path, and options controlling local saves/download hosts. The number of env vars is high because the skill supports integrations (OpenClaw, MCP, custom paths), but each variable has a clear purpose. No unrelated cloud or secret credentials (AWS, GCP, etc.) are requested.
Persistence & Privilege
The skill is not always-enabled (always: false) and does not request elevated platform-wide privileges. It writes to its own configured directories (last-render metadata, ~/Downloads/sogni by default) and can be configured to disable local saves. It does not attempt to modify other skills or global agent settings.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install sogni-agent - After installation, invoke the skill by name or use
/sogni-agent - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.5.16
Harden ClawHub packaging, remove runtime package.json dependency, and restrict MCP auto-download hosts.
v1.5.15
Align installed OpenClaw plugin behavior with SKILL.md and improve LTX 2.3 prompting guidance.
v1.5.14
Update Claude/OpenClaw skill routing so HD/4K video requests prefer the LTX family.
v1.5.13
Upgrade sogni-client-wrapper to 1.5.2; add ltx2.3, ia2v/a2v workflows, and API key auth support.
v1.5.11
v1.5.11 declare credential env/config access metadata for ClawHub review
v1.5.10
v1.5.10 scanner-safe process/env refactor and performance follow-up
v1.5.9
- Updated version to 1.5.9.
- General maintenance: updated dependencies in package.json and package-lock.json.
- Minor documentation and metadata changes in SKILL.md and openclaw.plugin.json.
v1.5.8
- Update Sogni credential link in setup instructions to https://app.sogni.ai/
- Minor README and documentation updates
v1.5.7
sogni-gen 1.5.7
- Updated skill version to 1.5.7 and related metadata.
- Minor maintenance; documentation (SKILL.md) refreshed with no user-visible behavioral changes.
- No CLI or API modifications.
v1.5.6
Fix README image not displaying on npmjs.com
v1.5.5
Patch release: republish under canonical slug.
v1.5.4
Declare configurable credential/filesystem paths and MCP write controls; align manifest metadata with runtime behavior.
v1.5.3
sogni-gen 1.5.3
- Added new CLI utilities: `--extract-last-frame` to extract the last frame from a video, and `--concat-videos` to concatenate video clips using a safe ffmpeg wrapper.
- Introduced `--list-media [type]` option to list recent inbound media (images, audio, all).
- Updated documentation to reflect new CLI options and usage instructions.
v1.5.2
sogni-gen v1.5.2
- Improved: Refined handling of arguments and edge cases in the main CLI.
- Fixed: Minor bug fixes in prompt processing.
- Maintenance: Dependency updates and small codebase cleanups.
v1.5.1
Version 1.5.1
- Adds support for v2v (video-to-video) workflows and LTX-2 models.
- New options for v2v and animate-replace workflows (e.g., --controlnet-name, --sam2-coordinates, --trim-end-frame, keyframe strengths).
- Updated CLI documentation with new flags, workflows, and models.
- Removes desktop-extension and test-related files for a leaner package.
- Minor documentation improvements and internal refactoring.
v1.4.0
- Added video generation support alongside image generation using Sogni AI's decentralized network.
- Expanded command-line options for video workflows, including text-to-video, image-to-video, and animation controls.
- Documented new and updated CLI flags for advanced video parameters (fps, duration, workflows, reference assets).
- Updated detailed installation and usage instructions covering both image and video creation.
- Listed default model IDs and settings for both images and videos, including configuration for OpenClaw integration.
Metadata
Frequently Asked Questions
What is Sogni Gen?
Generate images **and videos** using Sogni AI's decentralized network, with local credential/config files and optional local media inputs. Ask the agent to "... It is an AI Agent Skill for Claude Code / OpenClaw, with 955 downloads so far.
How do I install Sogni Gen?
Run "/install sogni-agent" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Sogni Gen free?
Yes, Sogni Gen is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Sogni Gen support?
Sogni Gen is cross-platform and runs anywhere OpenClaw / Claude Code is available (darwin, linux, win32).
Who created Sogni Gen?
It is built and maintained by Mauvis Ledford (@krunkosaurus); the current version is v1.5.16.
More Skills