The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (obsidian-exec.sh, obsidian-preflight.sh) and the obsidian CLI binary to interact with the vault. Scripts use Bash arrays to safely pass arguments, preventing shell injection vulnerabilities. The skill access is limited to the functionality of the Obsidian CLI and is consistent with the skill's purpose of vault management.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests data from external files (Obsidian notes) without sanitization or boundary markers. (1) Ingestion points: Vault content is read via safe_exec read and safe_exec search in scripts/obsidian-exec.sh. (2) Boundary markers: The scripts do not wrap the retrieved content in specific delimiters to warn the LLM against embedded instructions. (3) Capability inventory: The skill can read, write, and search files via the obsidian binary. (4) Sanitization: No filtering or escaping is applied to the content retrieved from vault notes.
[EXTERNAL_DOWNLOADS]: Error messages in the scripts suggest installing coreutils using brew install coreutils if the timeout command is missing. This is a manual remediation step for the user, not an automated download.

obsidian-cli