The Agent Skills Directory

[SAFE]: The skill is explicitly defined as read-only and is designed to diagnose, validate, and report on the health of a wiki structure without modifying files.
[CREDENTIALS_UNSAFE]: While the skill includes a sensitive-pattern check (as seen in references/doctor-checks.md and references/safety-rules.md), it contains explicit instructions and safety rules to avoid printing or exposing secret values, tokens, or passwords in its output. It treats these findings as cleanup signals rather than permission to display the content.
[COMMAND_EXECUTION]: The skill executes a local Python script (llm_wiki.py). The path resolution logic is restricted to the skill's own directory structure, and it includes a check to ensure the script exists before invocation. The discovery process for the 'vault root' is limited to specific numbered candidates and does not perform broad disk scanning.
[EXTERNAL_DOWNLOADS]: The skill references a command (npx skills add huajiexiewenfeng/obsidian-llm-wiki) as a recommendation if the runtime is missing. This targets the author's own repository and is a standard installation pattern for this ecosystem.

obsidian-wiki-doctor