tavily-search

PassAudited by ClawScan on May 10, 2026.

Overview

This skill appears to perform Tavily web search and URL extraction as advertised, but users should notice that it uses a Tavily API key, returns untrusted web content, and has a small metadata provenance mismatch.

This looks safe to install if you want Tavily-powered web search and extraction. Before using it, confirm the publisher/package identity, provide only a Tavily API key you are comfortable using here, and remember that web results should be treated as untrusted reference text rather than instructions.

Findings (3)

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 webpage or search result could contain text that tries to influence the agent if the agent treats retrieved content as instructions instead of reference material.

Why it was flagged

The skill prints raw content extracted from user-supplied URLs, which is expected for a web extraction tool but can include untrusted instructions or prompt-injection text from webpages.

Skill content
const content = String(r?.raw_content ?? "").trim(); ... console.log(content || "(no content extracted)");
Recommendation

Treat search results and extracted webpage text as untrusted data; do not let instructions inside retrieved pages override the user's request or safety rules.

What this means

Using the skill may consume the user's Tavily quota and sends the user's search queries or requested URLs to Tavily.

Why it was flagged

The skill requires a Tavily API key and the scripts use it to authenticate calls to Tavily's search and extract endpoints, which is purpose-aligned for this integration.

Skill content
Needs `TAVILY_API_KEY` from https://tavily.com
Recommendation

Use a dedicated Tavily key if possible, monitor usage, and avoid submitting sensitive queries or URLs unless you are comfortable sharing them with Tavily.

What this means

The skill's behavior is transparent in the included files, but the publisher/package identity is slightly inconsistent.

Why it was flagged

The embedded package metadata differs from the registry metadata shown for this evaluation, which lists a different owner ID and slug (`tavily-search1`). This is a provenance inconsistency, not evidence of malicious behavior.

Skill content
"ownerId": "kn7azq5e6sw0fbwwzdpcwvvjzd7z0x4z", "slug": "tavily-search"
Recommendation

Verify that this is the package and publisher you intended to install, especially if you expect an official Tavily-maintained skill.