study
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manages persistent state in a dedicated local directory (~/.study/), which is clearly documented and used solely for storing learning topics, records, and sandboxes.
- [SAFE]: Utilizes self-contained HTML templates for educational overviews and retrieval practice. The included JavaScript is restricted to local feedback collection via localStorage and browser-triggered downloads, with no external network activity or dynamic script execution from untrusted sources.
- [SAFE]: Implements a multi-tool ingestion model where other skills can append to the learning queue. Ingestion occurs at ~/.study/topics.md. The agent is instructed to treat this data as plain-text human context and avoid parsing machine-metadata, providing a boundary for potential indirect prompt injection. Capabilities are limited to local file operations within the study directory.
- [SAFE]: The documentation references opening generated HTML files using xdg-open or open and optionally POSTing feedback to a local server when the separate canvas skill is installed. These are intended behaviors for the skill's interactive learning and review workflows.
Audit Metadata