obsidian-search

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python script (find_similar.py) via the Bash tool to perform semantic search calculations. This script resides within the author's local workspace at /Users/rami/Documents/life-os. Standard utilities such as grep and glob are also used for keyword fallbacks and folder browsing.
  • [INDIRECT_PROMPT_INJECTION]: The skill reads the contents of files from the user's Obsidian vault, which constitutes an ingestion of untrusted data if the notes contain content clipped from external sources.
  • Ingestion points: The Read tool is used to access notes found via search in SKILL.md.
  • Capability inventory: The agent has access to Bash for command execution and Read for file system access.
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to distinguish note content from system instructions.
  • Sanitization: No content filtering or validation is performed on the note text prior to presentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 10:16 PM
Security Audit — agent-trust-hub — obsidian-search