The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the obsidian command-line tool to perform various vault operations, such as note creation, search, and property management.
[DYNAMIC_EXECUTION]: The obsidian eval command, documented for plugin development, allows the execution of arbitrary JavaScript within the Obsidian application context. This provides broad access to application internals and vault data.
[EXTERNAL_DOWNLOADS]: The skill suggests the installation of the mcp-obsidian-cli package via npm as an alternative if the native CLI tool is not present.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and processing Markdown content and YAML-based .base files from a user's vault. These files constitute untrusted input that could potentially contain malicious instructions designed to influence the agent's behavior.
Ingestion points: obsidian read, obsidian search, and processing of .base and .md files.
Boundary markers: The skill includes a general "human confirmation first" instruction but lacks specific delimiters or sanitization for vault content.
Capability inventory: File read/write, arbitrary JavaScript execution (obsidian eval), screenshot capture (obsidian dev:screenshot), and property modification across all reference scripts.
Sanitization: No explicit escaping, validation, or filtering of external content read from the vault is defined in the instructions.

obsidian