The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill installs obsidian-cli using a third-party Homebrew tap (yakitrak/yakitrak/obsidian-cli). Since the author and repository are not on the trusted list, there is a risk that the binary could contain malicious code or be updated without oversight.
[COMMAND_EXECUTION] (MEDIUM): The skill relies on executing shell commands via obsidian-cli to perform file operations like search, create, move, and delete. While intended for vault management, this provides a pathway for arbitrary file manipulation if the CLI tool or the inputs are compromised.
[DATA_EXPOSURE] (LOW): The skill explicitly reads ~/Library/Application Support/obsidian/obsidian.json. This is a sensitive configuration file containing metadata about the user's vaults and workspace state.
[INDIRECT_PROMPT_INJECTION] (LOW):
Ingestion points: Reads and searches the contents of markdown files (*.md) within vaults.
Boundary markers: No delimiters or warnings are provided to prevent the agent from interpreting instructions found inside the notes.
Capability inventory: The skill can create, move, and delete files via the CLI.
Sanitization: There is no evidence of sanitization for note content before it is processed by the agent.

obsidian