Wonda CLI — AI Content Creation from the Terminal
Generate images, video, music, and voice from one terminal command, then publish straight to TikTok, Instagram, LinkedIn, X, and Reddit — replacing manual multi-platform content workflows.
Docs disclose credential vault encryption, telemetry opt-out, and scoping, giving reasonable data-flow transparency; however, the skill encourages automated LinkedIn/X/Reddit operations, an antidetect stealth browser, bulk account creation and platform-detection evasion — ToS-violating gray-area risk. Passwords are decrypted with a server-side key and cookie cloud backup is on by default, neither least-privilege nor locally isolated. Deductions: stealth automation, default cloud backup of cookies, server-held decryption key.
Command structure is self-consistent, output conventions (JSON/quiet/jq) are uniform, and failure feedback is described in detail (fail-closed entitlement checks, unknown_project typo guard); but there is no test or execution evidence, heavy dependence on a paid external service, and complex wab update paths that cannot be statically validated. Deductions: no tests, reproducibility, or abnormal-path evidence.
Target scenario and audience (terminal content generation and social automation) are clear, with explicit paid-tier boundaries; but core function depends entirely on overseas services (wonda.sh, LinkedIn, X, Reddit, Meta), essentially unreachable from mainland-China networks, with no Chinese-language support noted. Deductions: poor environment fit and no declared non-fit environments.
Documentation is well structured with rich examples and versioned release channels; but the license is proprietary with no changelog, no contribution or maintenance-responsibility statement, and SKILL.md embeds excessive runtime detail. Deductions: opaque licensing, missing governance and change records.
Workflow examples are complete (generate→edit→publish pipeline) and output is script-ready; but value depends on a paid subscription ($19.99/49.99/mo) plus credits, with no execution evidence of output quality and no argued marginal value over official platform tools. Deductions: no verified results, insufficient cost/benefit evidence.
Only author claims and marketing material (README badges, screenshots); one 'verified 2026-07-07' claim is not independently checkable; no CI tests, no third-party execution evidence, no cross-source corroboration. Deductions: nearly all claims are unverifiable author assertions.
- Static review only; no commands were executed. All conclusions derive from documentation.
- The skill promotes platform automation with an antidetect stealth browser, likely violating LinkedIn/X/Reddit ToS — account bans and compliance risk.
- Core function depends entirely on overseas paid cloud services, likely unreachable from mainland-China networks, and requires a subscription from $19.99/mo.
- Platform cookies are cloud-backed-up by default and passwords are decrypted with a server-held key — assess sensitive-data custody risk yourself.
- License is proprietary; no source code is available to audit CLI behavior.
What does this skill do, and when should you use it?
Wonda CLI is a commercial content-creation toolkit designed for terminal-based AI agents, installable via npm, Homebrew, or signed desktop installers. It generates images, videos, music, and speech (including voice cloning from a 10s clip), plus TikTok/Reels-style editing operations like animated captions, trimming, background removal, and upscaling. It also covers social publishing and automation: posting to Instagram/TikTok and reading/writing LinkedIn, X/Twitter, and Reddit. The product is paid-only: free accounts get no product access, with WONDA at $19.99/mo and WONDA_PREMIUM at $49.99/mo. Stealth automation runs through the WAB antidetect browser, keeping platform cookies on-device.
The wonda CLI calls hosted generation models (image models like nano-banana-2, video like sora2, music like suno-music, TTS like minimax-speech), returning JSON job results filterable with --jq/--fields. It executes roughly 25 video/audio/image editing operations and writes processed files. It manages a media library, projects and org wallets, and an AES-256-GCM-encrypted credentials vault. It reads and writes LinkedIn/X/Reddit/Instagram through either a flat cookie store (--via cookies) or the WAB stealth Chromium (--via wab), including posting, commenting, liking, and DMs. It also publishes to Instagram/TikTok, scrapes competitor ads from the Meta Ads Library, and offers a cloud twin (premium) and a local relay.
- Indie creators/marketers: have an agent like Claude Code turn a product photo into a short video with music and burned-in TikTok captions, then publish it.
- Social media managers: bulk-read LinkedIn profiles and post engagement, auto-comment, like, and post.
- Growth teams: search the Meta Ads Library with `scrape ads` for competitor creative analysis.
- Content teams: clone a brand voice for multi-speaker TTS dialogue.
- Automation developers: wire wonda into pipelines using --quiet/--jq to chain generate→edit→publish.
What are this skill's strengths and limitations?
- End-to-end generate→edit→publish pipeline chainable in a single CLI
- Designed agent-first: JSON output, --quiet/--jq/--fields, auto-syncing built-in skill file
- WAB antidetect browser reduces platform anti-abuse risk on writes; cookies stay on-device
- Cross-platform (macOS/Linux/Windows, x64+ARM64) with multiple install channels
- Enterprise spend controls: org wallets, project tagging, usage reports
- Paid-only: no free tier; a positive credit balance doesn't substitute for a plan; voice cloning costs $1.50 each
- Proprietary closed source — no license shown on GitHub
- Generation depends on hosted cloud models; nothing runs locally
- Some risky operations (Reddit signup, email) are gated by per-account flags and can be disabled
- Cloned voices expire after 7 days of inactivity and must be re-cloned for another $1.50
- No visible test suite or independent benchmarks in the repo
How do you install this skill?
Install: npm i -g @degausai/wonda or brew tap degausai/tap && brew install wonda (needs Node >= v20). Authenticate: wonda auth login (browser) or set WONDA_API_KEY. Install the skill for your agent: wonda skill install -o . or npx skills add degausai/wonda; Claude Code can also use /plugin marketplace add degausai/wonda then /plugin install wonda@degausai. Note: you must subscribe to a paid plan at wonda.sh first, or all product commands return 403 paid_plan_required.
How do you use this skill?
Example: wonda generate image --model nano-banana-2 --prompt "..." --wait -o photo.png; wonda generate video --model sora2 --attach <media-id> --wait -o video.mp4. Then ask your agent something like "Use wonda to generate a product video of this image." Generation costs credits — check costs with wonda pricing estimate and balance with wonda balance. Before publishing, set up platform sessions via wonda wab login or wonda x auth set.