User and Vendor Awareness and Perception of the Agentic AI Skill Ecosystem
Security awareness and perspective of users and solution suppliers towards agentic AI skills running on new IoT platforms such as MoltBot, ClawBot and the Clawdhub skill ecosystem.
Why Study Agentic AI Skills on IoT?
A new generation of IoT devices ships with agentic AI capabilities: voice- and chat-driven assistants that can discover and install third-party "skills" in order to extend their behavior at runtime. Platforms such as MoltBot and ClawBot, together with skill marketplaces like Clawdhub, have lowered the barrier for independent developers to publish IoT-facing functionality to millions of end users.
How users and vendors actually perceive the security and privacy implications of these skills is still largely unknown. Our research approaches both sides of the relationship as open questions: we want to understand how end users think about the skills they install and use, and how solution suppliers think about the responsibility and risk that come with publishing them.
Take part in our survey
We are currently looking for participants to share their perception and awareness of agentic AI skills on IoT platforms. The survey takes only a few minutes and your answers directly support the research described on this page.
Methodology
The project combines qualitative, quantitative, and technical methods so that findings on one side can be validated on another.
Interviews
Semi-structured interviews with end users and solution suppliers are used to collect in-depth context on security awareness and perception. They let us explore how people reason about agentic AI skills, which assumptions they make, and where their mental models break down. Transcripts are coded thematically to surface recurring concerns, misconceptions, and blind spots.
Survey
A structured survey is used to collect wide demographic data across different user groups, so that the qualitative findings from the interviews can be generalized and compared across age, technical background, and usage patterns. This allows us to quantify how widespread specific beliefs and behaviors are in practice.
Clawdhub Skill Analysis
In parallel, we perform a technical analysis of skills published on the Clawdhub platform: permissions requested, data flows, update behavior, and the clarity of the privacy and security information presented to the user at install time.
Status
The project is ongoing. Interview and survey data collection is in progress. Findings will be published here as they become available.