The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides an interface to execute a wide range of terminal commands using the obsidian CLI, allowing the agent to manipulate the local file system and application state.
[REMOTE_CODE_EXECUTION]: The inclusion of the eval command in references/commands.md allows the agent to execute arbitrary JavaScript code within the Obsidian application's runtime environment. This is a critical risk as it could be used to perform unauthorized actions or exfiltrate data.
[EXTERNAL_DOWNLOADS]: Commands such as plugin:install and theme:install enable the agent to download and install external code from community repositories, which may not be vetted for security.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection attacks because it is designed to read and process content from user-controlled notes.
Ingestion points: Content is ingested through obsidian read, obsidian search, and obsidian daily:read commands.
Boundary markers: There are no instructions or delimiters provided to prevent the agent from following instructions embedded within the notes it reads.
Capability inventory: The agent possesses powerful capabilities including eval, delete, move, and plugin:install.
Sanitization: No evidence of content validation or sanitization exists to mitigate the risk of malicious instructions contained within processed data.

obsidian-cli