The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents an extensive range of commands to control the local Obsidian application via its CLI binary, including file management, workspace layout manipulation, and system settings control.
[REMOTE_CODE_EXECUTION]: Includes the obsidian eval command, which allows for the execution of arbitrary JavaScript code directly within the running Obsidian application's environment.
[REMOTE_CODE_EXECUTION]: Features commands such as obsidian plugin:install and obsidian theme:install that can download and enable third-party code from community repositories, facilitating potential remote code execution.
[DATA_EXFILTRATION]: Documents commands for obsidian sync and obsidian publish, which are designed to transmit vault content to external servers, as well as an obsidian web command that can open arbitrary URLs in a web viewer.
[PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection as it ingests untrusted data from the Obsidian vault without isolation or validation.
Ingestion points: Vault content enters the agent's context through commands such as obsidian read, obsidian daily:read, obsidian search, and obsidian template:read in SKILL.md.
Boundary markers: Absent; there are no delimiters or instructions to prevent the agent from executing commands found within the notes it processes.
Capability inventory: The skill provides access to JavaScript execution (eval), plugin management, and network-enabled sync/publish features.
Sanitization: Absent; the content retrieved from the vault is not sanitized or filtered before processing.
[COMMAND_EXECUTION]: Provides the obsidian dev:cdp command for raw Chrome DevTools Protocol interactions, allowing for deep inspection and manipulation of the application's internal state.

obsidian-cli