Back to skill

Security audit

openclaw-china-search-tips

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its advertised purpose (China-friendly search with multiple backends) but contains several mismatches and sloppy/odd code (undeclared env vars, an unexpected search endpoint, and a logic bug that can treat empty results as success).

Things to check before installing/use: - Do not export sensitive global credentials until you confirm the endpoints. The skill asks you to set VOLC_SEARCH_API_KEY / TAVILY_API_KEY / SEARCHAPI_API_KEY, but the registry metadata does not declare them — that discrepancy is sloppy and could cause surprise. - Inspect and verify the endpoints in the bundled code. The volcengine handler posts to https://open.feedcoopapi.com/search_api/web_search which is not an obvious volcengine domain; confirm this host is legitimate before sending an API key. - Be aware of a functional bug: try_volcengine returns an empty list on some successful HTTP responses, and the main search() treats any non-None return value (including empty list) as success. This may cause the skill to report success but return no results. - If you plan to use the multi-search fallback, review or install the multi_search_engine package separately and inspect its behavior. - Prefer creating dedicated, least-privileged API keys (not reuse highly privileged tokens) and test the skill in a sandbox environment first. If unsure, ask the publisher to update the registry metadata to declare required env vars and to explain the volcengine endpoint.

SkillSpector findings are pending for this release.