The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to manage the Obsidian application state and its cache.
The agent is directed to use pgrep -x Obsidian to check if the application is currently running before performing maintenance.
The skill uses rm -rf ~/Library/Application\ Support/obsidian/IndexedDB/ to clear the cache. While rm -rf is a powerful command, its use here is restricted to a specific, non-data cache directory and is the primary purpose of the utility.
Safety instructions are explicitly included, requiring the agent to use AskUserQuestion to confirm the application is closed and the user agrees to the deletion before execution.
[INDIRECT_PROMPT_INJECTION]: The skill contains a 'Generic feedback loop' section that instructs the agent on how to handle user modification requests.
The instructions tell the agent to distinguish between task-specific and reusable changes, potentially updating 'canonical Skills' or shared specifications based on user input.
This defines a framework for the agent to process untrusted user data to modify the skill's own rules or other files in the environment. However, this is presented as a standard architectural pattern for the 'lovstudio' skill set and does not include malicious override patterns.

lov-obsidian-reset-cache