Toggle

• Analyzed: May 12, 2026, 4:21 AM
• Findings: 1
• Version: 1.0.6
• Hash: 27e93a67dca4249d7ca8a78aeb2247a4f791451761f2889355c6545313e7e76b
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill collects extensive user activity data (work sessions, projects, focus scores, context switches) and transmits it to an external third-party service (ai-x.toggle.pro). It then writes this sensitive data into the agent's memory files (`.md` files) using a relative path (`../../memory`). While the `toggle.py` script wraps the JSON data in a markdown code block, the `SKILL.md` instructions explicitly direct the agent to interpret and act upon this data for various proactive behaviors (e.g., summarization, nudges, predictions). This creates a significant prompt injection vulnerability, as malicious or crafted data from the external service (or a compromised user account) could, if misinterpreted by the agent, lead to unauthorized actions. Additionally, the skill instructs the agent to create cron jobs, a powerful capability, even if the provided examples are for its intended function.