vault

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The skill’s /vault search and subsequent “Context reading” path uses grep to find matches in connected vault directories and then cat to read the matched .md files into the agent context; those vault files are outsider-authored content (the operating user did not author all text in the external vaults), so free-form prose from external vaults can be ingested at runtime.