obsidian-bases

Pass

Audited by Gen Agent Trust Hub on Aug 9, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill documents a DSL for Obsidian Bases that processes untrusted data (note frontmatter, file metadata) from the local vault. This creates a surface for indirect prompt injection, particularly through the html() function which allows rendering strings as HTML, potentially leading to XSS if note properties contain malicious scripts. The skill includes the escapeHTML() function to help mitigate these risks, although its implementation depends on the generated formula.\n
  • Ingestion points: Note frontmatter properties, file metadata (name, path, size, etc.), and file content accessed through the file(path) function.\n
  • Boundary markers: No specific delimiters or warnings for processed data are defined in the schema.\n
  • Capability inventory: Access to vault files via file(), HTML rendering via html(), regex matching via matches(), and various data manipulation functions defined in the references/FUNCTIONS_REFERENCE.md.\n
  • Sanitization: An escapeHTML() function is provided in the documentation but its use is not enforced by the schema or default examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 9, 2026, 10:24 PM
Security Audit — agent-trust-hub — obsidian-bases