The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents the use of the obsidian eval command, which allows the agent to execute arbitrary JavaScript code within the context of the running Obsidian application.
[COMMAND_EXECUTION]: The skill relies on an external command-line tool (obsidian) to perform a variety of operations including file management, plugin manipulation, and developer-level inspection.
[DATA_EXFILTRATION]: High-privilege commands are provided to capture screenshots (obsidian dev:screenshot), extract text from the application DOM (obsidian dev:dom), and read application console logs (obsidian dev:console), which can be used to access and exfiltrate sensitive data from the user's workspace.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its architecture:
Ingestion points: The skill retrieves data from the Obsidian vault and application UI using commands like obsidian read, obsidian search, and obsidian dev:dom (SKILL.md).
Boundary markers: There are no instructions or delimiters defined to prevent the agent from following instructions embedded in the retrieved vault content.
Capability inventory: The skill has extensive capabilities to modify the environment and execute code via obsidian create, obsidian append, and obsidian eval (SKILL.md).
Sanitization: The instructions do not describe any sanitization or validation of content fetched from the vault before it is processed by the agent.

obsidian-cli