The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like mkdir, cat, find, grep, mv, and rm with variables directly derived from user input.
Evidence: Commands in the 'Quick Commands' section use placeholders like {{Vault}}, {{folder}}, and {{filename}}, and search commands use keyword directly in grep and find arguments.
Risk: Although the skill instructs the agent to sanitize filenames, natural language processing can be bypassed. Malicious input containing shell metacharacters (e.g., ;, |, $(...)) could lead to arbitrary command execution on the host system.
[DATA_EXFILTRATION]: The skill grants the agent broad read and write access to the ~/Documents/Obsidian/ directory.
Evidence: Use of find ~/Documents/Obsidian -name "*.md" -type f | xargs grep -li "keyword" and grep -rli across multiple vaults.
Risk: Obsidian vaults often contain sensitive personal or professional data. The ability to search and read all files in this directory allows for the exposure of private information if the agent is manipulated into exfiltrating the content of retrieved notes.

obsidian