share-field-notes
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes community-contributed Markdown files which represent an external data ingestion surface. Malicious instructions could theoretically be embedded in shared notes.
- Ingestion points: The
um kb searchandum kb showcommands retrieve content from theknowledge/directory in the associated GitHub repository. - Boundary markers: The instructions explicitly warn the agent to treat findings as "strong hints, not gospel" and advise against running commands from notes blindly.
- Capability inventory: The skill can create new files (
um kb new), validate them (um kb check), and push changes to GitHub (um kb pr). - Sanitization: The
um kb checkutility is mandated to scan for secrets and unfilled templates before any external submission. - [EXTERNAL_DOWNLOADS]: The skill installs the
universal-moddertool from the author's GitHub repository and retrieves metadata from GitHub's raw content service. These operations target the vendor's own infrastructure and established content delivery networks.
Audit Metadata