NYC

obsidian-vault-context

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [Data Exposure] (LOW): Hardcoded absolute file paths (e.g., /Users/steve/Documents/Main) expose a specific local username and internal directory structure.
  • [Indirect Prompt Injection] (LOW): The skill is designed to ingest external research content and write it directly into markdown files without sanitization or boundary markers.
  • Ingestion points: Research capture template and existing note content read via read_file in SKILL.md.
  • Boundary markers: Absent; the skill uses direct string concatenation without delimiters for external data.
  • Capability inventory: File writing (write), text manipulation (search_replace), and CLI interaction (obsidian-cli).
  • Sanitization: Absent; the instructions assume external content is safe to append.
  • [Unverifiable Dependencies] (LOW): The skill relies on obsidian-cli, an external tool whose source and integrity are not specified or verified within the skill context.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:49 PM