Paw Chat

WarnAudited by ClawScan on May 10, 2026.

Overview

Paw Chat appears to be a coherent OpenClaw web chat frontend, but it stores a gateway token in the browser and appears to render chat Markdown/HTML without sanitization, which could let malicious content run code in the page.

Treat this as a Review item rather than proven malware. Before installing, verify the source/package contents, use only a trusted Gateway URL, prefer a revocable/scoped token, and patch or confirm sanitization of rendered Markdown/HTML so malicious chat content cannot run in the browser.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A malicious assistant response, session history item, or tool output could potentially inject browser-executed code into the Paw page and then access the saved gateway token or interact with the gateway as the user.

Why it was flagged

Chat content is converted to HTML using marked, while custom image/link renderers interpolate href/text into HTML without visible sanitization or URL-scheme restrictions.

Skill content
return `<img loading="lazy" data-src="${href}" alt="${escHtml(alt)}" class="chat-img lazy-img" onclick="showImageLightbox(this.src)">`; ... return `<a href="${href}" target="_blank" rel="noopener">${text}</a>`; ... return marked.parse(text);
Recommendation

Sanitize all rendered Markdown/HTML with a vetted sanitizer, escape attributes and link text, restrict URL schemes, and add a strict Content Security Policy before using this with untrusted content.

What this means

The token is expected for connecting to the Gateway, but anyone with access to that browser profile—or injected page script—may be able to reuse it for OpenClaw actions.

Why it was flagged

The frontend persistently stores the OpenClaw Gateway token in browser localStorage.

Skill content
localStorage.setItem('oc-chat-cfg', JSON.stringify({ url: cfgUrl.value, token: cfgToken.value, session: cfgSession.value }))
Recommendation

Use a revocable or scoped Gateway token if available, avoid shared browsers, clear localStorage when done, and ensure the registry metadata clearly declares the credential use.

What this means

Chat messages, uploaded images, and the token will be sent to the Gateway URL the user configures; a wrong or untrusted URL would receive that data.

Why it was flagged

The skill discloses a user-configured WebSocket connection to an OpenClaw Gateway and support for uploading images.

Skill content
WebSocket 连接:直接连接到 OpenClaw Gateway ... 图片支持:支持粘贴、拖拽、上传图片 ... Gateway URL ... Token
Recommendation

Verify the Gateway host before connecting, prefer wss:// or localhost for sensitive use, and avoid uploading private files to untrusted gateways.

What this means

The reviewed artifact set appears incomplete, so installation may fail or users may be pushed toward unreviewed external package paths.

Why it was flagged

The install script requires assets/logo.jpg, but the provided file manifest does not include logo.jpg; package.json also references CLI files not present in the manifest.

Skill content
set -e ... cp "$ASSETS_DIR/logo.jpg" "$UIROOT/"
Recommendation

Ship all referenced assets and CLI files with the skill, or update the docs/scripts to match the reviewed package contents.