Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Link Transcriber Skill Public

v0.1.0

Use this skill when a user wants to submit a Douyin or Xiaohongshu link to the linkTranscriber transcription API, optionally provide cookie when available, w...

0· 35·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The skill's name, description, SKILL.md workflow, and the included example script all align: accept a Douyin/Xiaohongshu link (and optional cookie), create a transcription task, poll it, call a summaries API, and return summary_markdown. Asking for a cookie is coherent with the purpose (some APIs require cookies to access private content). However, the default API endpoint is a raw IP (http://139.196.124.192/linktranscriber-api) rather than a documented domain, which is unusual for a public integration and worth questioning.
!
Instruction Scope
The SKILL.md and example script instruct the agent to collect a user-provided cookie and POST it (if provided) together with the link to the upstream service. The default transport is plain HTTP to a numeric IP (no TLS), so cookies and link data would be transmitted unencrypted. The instructions do not ask the agent to read local browser cookie stores or system files, but they do explicitly instruct the agent to request and forward potentially sensitive cookie strings to an external host — this is within the stated purpose but increases privacy/risk and should be made explicit to users.
Install Mechanism
No install spec is present (instruction-only distribution), which reduces installation-time risk. There is a single helper script (scripts/call_service_example.py) included for local testing; it uses only Python stdlib HTTP calls. Nothing in the package attempts to install third-party binaries or download/extract code at install time.
!
Credentials
The skill declares no required environment variables, but the example script honors LINK_SKILL_API_BASE_URL and optional LINK_SKILL_SUMMARY_PROVIDER_ID and LINK_SKILL_SUMMARY_MODEL_NAME env vars. SKILL.md generically mentions 'environment explicitly overrides it' for the base URL but does not name these env vars in its Required Inputs section. This is an inconsistency in metadata. More importantly, the skill will transmit user-supplied cookies to the upstream service; because the default endpoint is an HTTP numeric IP, this may expose sensitive cookies in plaintext on the network unless you override the base URL to a trusted HTTPS endpoint.
Persistence & Privilege
The skill does not request persistent or elevated privileges (always:false). It does not modify other skills or agent-wide configuration. It will not autonomously write credentials to the agent. Autonomous invocation is allowed (platform default) but not combined with other high privileges here.
What to consider before installing
This skill does what it says (transcribe and summarize Douyin/Xiaohongshu links), but exercise caution before installing or using it: 1) Do not paste or share sensitive authentication cookies unless you trust the upstream service operator. The default endpoint is a raw IP over HTTP (http://139.196.124.192) — that means data (including cookies) would be sent unencrypted. 2) Ask the skill author for a domain name, HTTPS support, and who runs the server (privacy/ownership). 3) If you must try it, consider overriding LINK_SKILL_API_BASE_URL to a trusted HTTPS proxy/service you control, or avoid providing cookies and test only with public links. 4) The package metadata omits explicit env var names used by the script (LINK_SKILL_API_BASE_URL, LINK_SKILL_SUMMARY_PROVIDER_ID, LINK_SKILL_SUMMARY_MODEL_NAME); confirm these before relying on overrides. If you cannot verify the server/operator or reject sending cookies unencrypted, do not provide sensitive cookie strings and treat the skill as untrusted.

Like a lobster shell, security has layers — review code before you run it.

latestvk9759abdvkfqw91gj3g8p074px8401dr

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments