The Agent Skills Directory

[SAFE]: The skill utilizes a set of domain-specific tools (obsidian_get_broken_links, obsidian_list_notes, obsidian_get_backlinks, obsidian_replace_in_note, obsidian_rag_query) to manage vault metadata and link structure. Its operations are restricted to the vault environment provided by the platform configuration.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes content from user-controlled notes (titles and text) during link audits and semantic suggestion phases.
Ingestion points: Reads note content and titles via obsidian_list_notes and obsidian_rag_query in SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content within the notes during processing.
Capability inventory: The skill has the capability to write to the file system using obsidian_replace_in_note in SKILL.md.
Sanitization: No specific sanitization or filtering of note content is defined. However, the skill provides a critical safety instruction requiring per-change user confirmation before any edits are performed, mitigating the risk of autonomous unauthorized modifications.

link-audit