Is mailerlite safe?
https://clawhub.ai/byungkyu/mailerlite
This is a legitimate MailerLite API integration skill that provides comprehensive documentation for managing email subscribers, campaigns, groups, and automations through the Maton OAuth proxy. The primary risk is trust delegation to the maton.ai third-party service which proxies all API traffic and OAuth tokens. No malicious code, prompt injection, or data exfiltration attempts were detected in the skill itself.
Category Scores
Findings (7)
MEDIUM Third-party API proxy trust delegation -15 ▶
All MailerLite API traffic is routed through gateway.maton.ai, which acts as an OAuth proxy. The user's MATON_API_KEY and all subscriber data pass through this third-party service. While this is the intended design for managed OAuth, it means maton.ai has full visibility into all API operations and data.
MEDIUM Cross-skill reference to api-gateway -10 ▶
The skill description contains a direct reference to another skill: 'For other third party apps, use the api-gateway skill (https://clawhub.ai/byungkyu/api-gateway)'. This could prompt the agent to install or invoke a second skill, expanding the attack surface. While not malicious by itself, it creates a skill-chaining pathway.
LOW Destructive operations available without guardrails -15 ▶
The skill documents destructive operations including deleting subscribers, forgetting subscribers (GDPR), deleting campaigns, deleting groups, and scheduling instant campaign delivery. An agent following these instructions without user confirmation could cause data loss or send unintended emails.
LOW Webhook creation could enable data exfiltration -15 ▶
The webhook API allows creating webhooks pointing to arbitrary URLs. A manipulated agent could create a webhook to an attacker-controlled endpoint, causing MailerLite to send subscriber event data (creates, updates, etc.) to that endpoint.
LOW Environment variable exposure in code examples -7 ▶
Code examples use os.environ['MATON_API_KEY'] directly in HTTP requests. While standard practice, if the agent logs or displays these requests, the API key could be exposed. The troubleshooting section even suggests running 'echo $MATON_API_KEY'.
INFO Skill uses emoji in metadata 0 ▶
The skill metadata includes an emoji field (brain emoji). This is cosmetic and not a security concern.
INFO Platform reads sensitive files during install -15 ▶
The filesystem monitoring captured reads of .env, .aws/credentials, and various config files during installation. These are attributed to the host platform (openclaw) initialization, not to the skill itself. The skill repository contains no executable install scripts.