Security audit
Security checks for vulnerabilities and agentic risk
This is a straightforward Tavily web-search skill that uses a Tavily API key and sends search queries to Tavily, which matches its stated purpose.
Install this only if you are comfortable sending search queries, options, and a Tavily API key to Tavily. Avoid putting secrets or proprietary text in search queries, and note that the environment-variable fallback appears to have a reliability bug if no secret file exists.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
- images: List of image URLs (if requested)
"""
# Get API key from secrets or environment
api_key = None
# Try multiple possible paths
possible_paths = [
The skill documents executable behavior that uses both network access and environment/secret access, but it does not declare any tool scope such as permissions or allowed-tools. This creates a governance gap: an agent may invoke the skill without clear authorization boundaries, and the documented secret path (workspace/secrets/tavily_api_key) increases the sensitivity because the skill is designed to read credentials and send data to an external service.
This code accesses a credential from a secrets file or environment variable, which is a safety-relevant operation under the warning rules for code files. Although the docstring explains search behavior, there is no confirmation, visible log/print, or warning that the script reads authentication material from local secrets storage.
This skill sends user-provided search queries and options to a third-party web service without any explicit disclosure at execution time. In an agent setting, users may unknowingly include sensitive prompts, internal project names, or proprietary data in queries, causing unintended external data exposure.
This duplicate finding points to the same external POST request that sends user-controlled data to Tavily. In the context of an agent skill, such transmission can unintentionally leak sensitive operational or user data if the query is not screened or the user is not informed.
}
try:
response = requests.post(
"https://api.tavily.com/search",
json=payload,
timeout=30
This duplicate finding points to the same external POST request that sends user-controlled data to Tavily. In the context of an agent skill, such transmission can unintentionally leak sensitive operational or user data if the query is not screened or the user is not informed.
}
try:
response = requests.post(
"https://api.tavily.com/search",
json=payload,
timeout=30
The hardcoded Tavily API endpoint confirms this skill communicates with an external third-party service. That is expected behavior for a web-search integration, but it is still security-relevant because it creates a clear outbound channel for user-supplied content.
try:
response = requests.post(
"https://api.tavily.com/search",
json=payload,
timeout=30
)