atomic-note-audit

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses subprocess.run() to interface with the Obsidian CLI and local agent runners. This is necessary for its functionality and is implemented with safeguards like Base64 transport encoding to prevent accidental shell command injection.
  • [DYNAMIC_EXECUTION]: The skill utilizes the Obsidian CLI's eval command to execute JavaScript within the running Obsidian application. This is used for reading and processing notes within the app context, which is more reliable than direct filesystem access for live vaults.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-generated notes through an LLM. It employs boundary markers (NOTE_CONTENT_START/NOTE_CONTENT_END) and specific system instructions to the model to treat note content as data rather than instructions, reducing the risk of indirect prompt injection.
  • [OBFUSCATION]: Base64 encoding is used in scripts/obsidian_adapter.py and scripts/publish_audit_report.py. This is a technical implementation detail designed to ensure reliable transport of complex strings (like LaTeX or JavaScript code) through CLI arguments, not a malicious attempt to hide behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 08:17 PM