Skill flagged — suspicious patterns detected

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

patent-search

v1.0.9

Search, view and analyze patents directly in OpenClaw. Supports global patent databases with intelligent search and analysis features.

0· 250·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The SKILL.md and the Python code clearly require an API token for https://www.9235.net/api (PATENT_API_TOKEN or OpenClaw config key). However the registry metadata at the top of the report lists "Required env vars: none" and the skill was described as instruction-only; that contradicts the files included (multiple Python modules). Requiring a 9235.net token is coherent with a patent-search skill, but the metadata vs. bundle contents disagree.
!
Instruction Scope
Runtime instructions and code limit network activity to the third‑party patent API (9235.net) and local skill config.json. The code looks focused on search/detail/analysis and uses the token for API calls. However SKILL.md contains a pre-scan flag for unicode-control-chars (prompt‑injection pattern) — SKILL.md itself may include stealth characters that could attempt to manipulate an LLM or evaluation. Also some files call helper that resolves the token (patent_token.py); you should inspect that file to confirm it doesn't read unrelated host files or exec arbitrary commands.
Install Mechanism
No install spec is present (instruction-only install), which is lower risk. The package includes multiple Python files that will be installed as part of the skill bundle; there is no evidence of external downloads or archive extraction. This is reasonable for a Python skill, but verify local execution permissions before running.
!
Credentials
The only required secret used by code and SKILL.md is PATENT_API_TOKEN (and optionally the OpenClaw config key). That credential is proportionate to the stated purpose. The concern is the mismatch between the registry metadata (which claimed no required env vars) and the explicit credential requirement in SKILL.md and the code; this inconsistency could cause accidental deployment without expected secrets or produce confusing behavior. Confirm how the platform supplies the token and inspect patent_token.py to see how it resolves OpenClaw config keys.
Persistence & Privilege
The skill is not always-enabled, does not request system-wide persistence or elevated privileges, and does not declare modifications to other skills. Autonomous model invocation is allowed by default (normal).
Scan Findings in Context
[unicode-control-chars] unexpected: A pre-scan detected unicode control (zero-width / bidi) characters in SKILL.md. That is not expected for a patent-search skill and could be used to perform prompt injection or to hide content. SKILL.md should be inspected/cleaned before trusting its instructions.
What to consider before installing
This package largely looks like a legitimate client for the 9235.net patent API and needs a PATENT_API_TOKEN. However: 1) Registry metadata claims no env vars while SKILL.md and the code require PATENT_API_TOKEN — confirm which is authoritative before installing. 2) Examine patent_token.py and check_env.py to ensure they only read the declared environment variable or call the OpenClaw CLI and do not open unrelated host config files or exec arbitrary commands. 3) Open SKILL.md in a hex-capable editor and remove/verify any zero-width or bidi-control characters flagged by the scan. 4) Because the skill makes network calls, run it in a sandbox or with a token that has limited/replaceable permissions while you evaluate. 5) If you will supply a real API token, confirm the provider's terms and rotate the token after testing. If you want, provide the contents of patent_token.py and check_env.py and I can review them for any code that reads unexpected files or performs subprocess/network calls beyond the declared API.

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

latestvk9714794fvczzx3f4escwpbsrh84fh31

License

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

Comments