The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from markdown files across the user's vault to perform tag audits and normalization.
Ingestion points: Reads all markdown files within $VAULT_PATH (excluding specific directories) to extract the tags field from YAML frontmatter and analyze page content.
Boundary markers: No explicit boundary markers or instructions to disregard embedded content are used when processing the files.
Capability inventory: The skill has file-write permissions (updating frontmatter, appending to the taxonomy file) and can execute shell commands (obsidian-wiki, qmd).
Sanitization: The skill does not specify any sanitization or validation logic for the tags extracted from potentially untrusted user files before they are written back to the vault or used in CLI commands.
[COMMAND_EXECUTION]: The skill executes local command-line tools to synchronize state and refresh search indices.
Tool execution: Invokes obsidian-wiki memory sync with dynamic arguments (e.g., tags_normalized, pages_modified) to track wiki changes.
Index management: Executes qmd update, qmd embed, and qmd ls/get to maintain the search index. It uses shell parameter expansion ${QMD_CLI:-qmd} to determine the binary path, which is a standard pattern for environment-provided tools.

tag-taxonomy