t-web-searcher

AI-optimized web search via Tavily API. Returns concise, relevant results for AI agents.

This skill appears to implement a Tavily web-search client and only needs your TAVILY_API_KEY and Node to run — that part is coherent. However, the search script trusts the API's meta.formatFile value and will attempt to import and execute that module at runtime. That means a response from the Tavily API could cause arbitrary code to run in your agent process (or attempt to import local files). Before installing: 1) verify the Tavily domain (api.tavily-search.com) and that you trust the service and owner; 2) consider the metadata mismatch in _meta.json as a minor red flag and confirm provenance; 3) if you must use this skill, run it in an isolated environment (container) and/or audit/patch scripts/search.mjs to only allow known local formatter modules (e.g., whitelist './formatter/*.mjs') rather than importing arbitrary URLs; 4) issue a least-privilege API key to Tavily that can be revoked quickly. If you cannot validate the service or patch the dynamic import behavior, treat this skill as risky and avoid providing sensitive credentials.

Type: OpenClaw Skill Name: t-web-searcher Version: 1.0.0 The skill contains a significant Remote Code Execution (RCE) vulnerability in `scripts/search.mjs`, where it dynamically imports a JavaScript module using a path derived from the remote API response (`data.meta.formatFile`). Additionally, both `scripts/search.mjs` and `scripts/extract.mjs` employ minor obfuscation by using indirect references to `process.env` (via `process_t` and `key_t` variables) to access the `TAVILY_API_KEY`, a technique often used to evade simple static analysis tools.