Tavily Search Skill

Web search via Tavily API (alternative to Brave). Use when the user asks to search the web / look up sources / find links and Brave web_search is unavailable...

Security Scan

The skill's files and runtime instructions are consistent with a simple Tavily web-search integration; the main issues are minor metadata omissions and an odd install instruction, not evidence of malicious behavior.

Purpose & Capability

The skill's name/description (Tavily web search) matches what the scripts do: calling https://api.tavily.com/search and https://api.tavily.com/usage. It legitimately requires a Tavily API key. However, the registry metadata does not declare the API key or any required env var even though SKILL.md and search.sh require TAVILY_API_KEY or an apikey file — this mismatch is a packaging/metadata oversight.

Instruction Scope

SKILL.md instructs the agent/user to git clone the repo, request the user's Tavily API key, store it in a local 'apikey' file (or set TAVILY_API_KEY), and run ./search.sh. Runtime steps and files referenced are limited to the skill directory (apikey, blocklist files, search.sh). The instructions do not request unrelated system files, other credentials, or exfiltration to unexpected endpoints.

Install Mechanism

There is no formal install spec (instruction-only), and SKILL.md tells the user to git clone the GitHub repo. The code bundle also includes the same scripts, which makes the clone instruction redundant/odd but not dangerous. The GitHub URL is a normal source; no downloads from untrusted IPs or extract steps are used.

Credentials

The only credential used is the Tavily API key (TAVILY_API_KEY or apikey file). That is proportional to a web-search integration. The concern is that the registry metadata did not declare the need for this credential; the script also supports reading the same key from an env var or a file, and it sends the key as a Bearer token to api.tavily.com as expected.

Persistence & Privilege

The skill does not request persistent elevated privileges, does not set always:true, and only writes a local 'apikey' file (SKILL.md tells user to create it). It does not modify other skills or global agent configuration.

Assessment

This skill appears to do what it claims: call the Tavily search API and optionally filter results with a local blocklist. Before installing, verify the GitHub repo URL is the intended source and review the repo contents. Be aware the skill asks you to provide your Tavily API key and suggests storing it in a local file named 'apikey' (it sets chmod 600 in instructions). If you prefer, set TAVILY_API_KEY in your environment instead of writing a file. The main oddities are: the registry metadata doesn't list the required API key and SKILL.md asks you to git clone even though the package contains the scripts — these are packaging issues rather than security red flags. Run the script in an isolated environment if you are unsure, and verify it only contacts api.tavily.com (no other remote endpoints are present in the files).

OpenClaw Skill for performing high quality web searches using the Tavily API.

Directory Structure

tavily-search-skill/
├── SKILL.md                    ← must stay at root
├── search.sh                   ← must stay at root (entry script)
├── apikey                      ← API Key file (user creates this)
├── _meta.json
└── blocklist/
    ├── blocklist.json         ← blacklist config
    └── filter_blocklist.py     ← filter script

Agent Workflow

When instructed to install this skill, follow these steps in order:

Step 1: Install the skill

git clone https://github.com/JayeGT002/Tavily-Search-Skill.git tavily-search-skill
cd tavily-search-skill

Step 2: Request API Key from user

Ask the user: "请提供你的 Tavily API Key，用于配置搜索功能。获取地址：https://app.tavily.com/api-keys"

Wait for the user's API Key reply before proceeding.

Step 3: Create apikey file

echo "USER_PROVIDED_API_KEY" > apikey
chmod 600 apikey

Step 4: Verify installation

./search.sh "test"

A successful response contains search results and quota info.

---

Blacklist Management

The blacklist filters out low quality or untrusted sources from search results.

View current blacklist

cat blocklist/blocklist.json

Add a domain to blacklist

When the user says "block [domain]", update blocklist/blocklist.json:

Root domains automatically match all subdomains. Example: adding csdn.net also blocks blog.csdn.net, download.csdn.net, etc.

Filter feedback

When results are filtered, a message is written to stderr (not visible in normal output). Check stderr if you suspect filtering is silently removing results.

---

Usage

Basic Search

./search.sh "search query"

Specify Result Count

./search.sh "query" 10

Include Images

./search.sh "query" 5 true

Dependencies

• curl
• jq

Install if missing:

• Ubuntu/Debian: sudo apt-get install curl jq
• macOS: brew install curl jq
• Alpine: apk add curl jq