The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation describes the use of an external command-line tool, obsidian, to perform all vault operations. The agent is instructed to execute various subcommands through this CLI.
[REMOTE_CODE_EXECUTION]: The obsidian eval command allows for the execution of arbitrary JavaScript code within the runtime environment of the Obsidian application. This provides a mechanism for running unvetted code that can access internal app APIs and data.
[DATA_EXFILTRATION]: The skill provides several commands that expose sensitive user information, including obsidian read (file content), obsidian search (search results), obsidian dev:screenshot (visual workspace data), and obsidian dev:console (application logs).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from the user's vault that could influence agent behavior.
Ingestion points: Note content retrieved via obsidian read and obsidian search results.
Boundary markers: The instructions do not define delimiters or specific safety markers for distinguishing file content from agent instructions.
Capability inventory: The skill has access to file system operations, arbitrary JavaScript execution via eval, and UI inspection/capture tools.
Sanitization: There is no mention of sanitizing or escaping the data read from the vault before it is interpreted by the agent.

obsidian-cli