wk-silverbullet
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process content from SilverBullet pages, which constitutes an ingestion point for untrusted data. Ingestion points include the window.client.space.readPage API and browser automation tools used for DOM inspection. There are no explicit boundary markers or instructions to isolate or ignore embedded commands within the processed pages. The capability inventory includes the Bash tool and browser-based execution tools (browser_evaluate, browser_click), which could be leveraged if an injection occurs. No sanitization or data validation logic is specified for the page content handled by the agent.- [DYNAMIC_EXECUTION]: The skill templates include instructions for generating dynamic JavaScript onclick handlers. These templates use String.fromCharCode to programmatically assemble strings, which is a method used here to bypass SilverBullet's internal markdown parser limitations regarding double-quote characters in HTML attributes.- [EXTERNAL_DOWNLOADS]: The SKILL.md file provides a docker-compose configuration that references the SilverBullet container image. The image ghcr.io/silverbulletmd/silverbullet:2.8.1 targets a well-known service and the official repository for the software, representing a standard and safe deployment practice.
Audit Metadata