The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a variety of shell commands through the obsidian CLI tool to manage local files and vaults. Evidence: Numerous obsidian CLI invocations across SKILL.md (e.g., obsidian create, obsidian search, obsidian daily:append).
[REMOTE_CODE_EXECUTION]: The skill explicitly documents and encourages the use of an eval command that runs arbitrary JavaScript within the active Obsidian application context. This provides a direct path for executing unverified logic that can bypass typical file-system restrictions by operating within the application's memory space. Evidence: obsidian eval code="app.vault.getFiles().length" in SKILL.md.
[DATA_EXFILTRATION]: Multiple commands facilitate the exposure of private vault data to the agent's context or the system clipboard. This includes capturing screenshots of the application, reading console logs, and inspecting the DOM, which may contain sensitive personal information or credentials. Evidence: obsidian dev:screenshot, obsidian dev:dom, and obsidian dev:console in SKILL.md.
[COMMAND_EXECUTION]: The skill documentation encourages use of the --copy flag, which silently moves command output (potentially containing vault content) to the system clipboard, increasing the risk of accidental data exposure. Evidence: 'Use --copy on any command to copy output to clipboard' in SKILL.md.

obsidian-cli