Native package
ClawHub and OpenClaw
The published manifest, folder layout, environment variables, and install paths are designed for ClawHub and OpenClaw-style skill loaders.
Start with the native ClawHub package, connect one API key, and let your agent submit, poll, and return watermark-removal jobs. For other runtimes, use the portable Node command or integrate the same HTTP API directly.
01 · Choose the integration layer
The watermark workflow is reusable, but “universal” means different things at the package, command, and API layers. Choose the narrowest layer your runtime can support reliably.
Native package
The published manifest, folder layout, environment variables, and install paths are designed for ClawHub and OpenClaw-style skill loaders.
Portable command layer
The bundled .mjs script accepts command-line arguments, reads API_KEY, and prints structured JSON. Another agent can call it when its runtime supports local Node commands.
Runtime-neutral contract
A custom agent, backend, or automation can skip the skill package and call the authenticated watermark endpoints directly using multipart files or image URLs.
02 · Complete the first run
The bundled command keeps the agent handoff predictable: one credential, one input source, explicit async state, and an optional local artifact.
Use ClawHub for the native path, or extract the versioned zip into a compatible local skill directory.
Copy the raw sk-... secret shown once by your dashboard and expose it as API_KEY. Do not hash it yourself. API_BASE_URL is optional and only applies to a compatible server.
Submit one local file or one HTTP(S) image URL. The script returns structured JSON with status, task_id, output_url, and a suggested next action.
If the 30-second wait window ends first, poll the task. Add --download true to save a completed image under .openclaw-artifacts/remove-watermark/.
03 · Understand the system boundary
The skill is a client of the existing Airemovewatermark system. It does not run the AI model locally, bypass credits, or keep a permanent result archive.
A local file is uploaded to the Airemovewatermark API. A remote URL is fetched by the service. Use only images you own or are authorized to edit.
Tasks can move through queued, processing, finalizing, succeeded, or failed. Inspect the JSON code and task status instead of assuming one request always finishes the job.
The default response exposes a temporary output URL. Local download is opt-in, and service-side inputs and outputs are scheduled for cleanup after 48 hours.
04 · Pick a setup path
ClawHub is the recommended native path. The versioned archive and public API remain available when you need tighter control over installation or orchestration.
Best when your runtime already understands ClawHub skills and you want the shortest supported install path.
Open ClawHubBest for a local OpenClaw-style skill folder or a controlled deployment that pins version 0.1.7.
Download ZipBest for another agent framework, backend, or automation that should own its HTTP requests, polling, and result storage.
Read API DocsThese commands pin the published 0.1.7 archive. Review a new version before changing the URL in an automated deployment.
mkdir -p ~/.openclaw/skills && \
curl -L https://assets.airemovewatermark.net/skills/remove-watermark-skill-0.1.7.zip -o /tmp/remove-watermark-skill-0.1.7.zip && \
unzip -o /tmp/remove-watermark-skill-0.1.7.zip -d ~/.openclaw/skillsNew-Item -ItemType Directory -Force "$HOME\.openclaw\skills" | Out-Null
Invoke-WebRequest -Uri "https://assets.airemovewatermark.net/skills/remove-watermark-skill-0.1.7.zip" -OutFile "$env:TEMP\remove-watermark-skill-0.1.7.zip"
Expand-Archive -Path "$env:TEMP\remove-watermark-skill-0.1.7.zip" -DestinationPath "$HOME\.openclaw\skills" -Force05 · Resolve the handoff
No. The published package is native to ClawHub and OpenClaw-style skill loaders. Its bundled Node script is portable to runtimes that can execute local commands, while the HTTP API is the most general integration layer for other platforms.
The remove command accepts either one local JPG, PNG, or WebP file or one HTTP(S) image URL. Current public limits are 10MB per image and a maximum edge of 4096px.
No. The skill reads the file and uploads it to the remote API for processing. Inputs and outputs remain available for up to 48 hours, so download completed work promptly.
The --wait true option polls for about 30 seconds. If the job is still queued, processing, or finalizing, run the returned task command again with its task_id.
Continue with the layer you need
Use ClawHub for the shortest path. Use the docs when your system needs to own authentication, polling, retries, or result storage.