The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Python scripts (claude-obsidian.py and wiki-mode.py) to manage vault configuration. All write operations follow a strict 'plan-then-apply' transaction model requiring user review and SHA256 verification, which prevents arbitrary or silent modifications.
[DATA_EXPOSURE]: The skill reads .vault-meta/mode.json to determine the filing methodology. This is a local configuration file specific to the vault's organization and does not contain sensitive credentials or personally identifiable information.
[REMOTE_CODE_EXECUTION]: No remote code execution or external downloads were detected. All scripts and templates are referenced locally within the project structure using absolute paths derived from the installation root.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied vault names and note titles. While these are ingestion points, the implementation uses a dedicated helper script (wiki-mode.py) that sanitizes inputs and confines paths, mitigating potential injection risks.

wiki-mode