The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides an eval command that executes arbitrary JavaScript within the Obsidian application environment. It also includes advanced developer commands such as dev:cdp and dev:debug for interacting with the Chrome DevTools Protocol, which allows for direct manipulation of the application's internal state and environment.\n
Evidence: SKILL.md (eval, dev:cdp, dev:debug commands)\n- [REMOTE_CODE_EXECUTION]: The plugin:install command allows the agent to download and install community plugins from external sources. This functionality introduces third-party code into the user's environment at the agent's discretion.\n
Evidence: SKILL.md (plugin:install command)\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from the user's Obsidian vault. Malicious instructions stored in notes could be executed using the skill's powerful capabilities.\n
Ingestion points: SKILL.md (read, search, daily:read, and property:read commands)\n
Boundary markers: Absent; no delimiters or instructions to ignore embedded commands are present in the documentation or script wrappers.\n
Capability inventory: scripts/obs.sh and SKILL.md (eval, delete, create, append, and publish commands).\n
Sanitization: Absent; content retrieved from the vault is processed without validation or escaping.\n- [COMMAND_EXECUTION]: Installation procedures involve modifying shell configuration files (.bashrc, .zprofile, and PowerShell profiles) to alter the system PATH. This persistent modification of the user's shell environment is documented for the user to perform manually.\n
Evidence: README.md (manual setup sections for macOS, Windows, and Linux)\n- [EXTERNAL_DOWNLOADS]: Fetches and installs the CLI tool from the author's official Homebrew tap.\n
Evidence: README.md (brew tap alexanderkinging/tap)

obsidian-cli-official