The Agent Skills Directory

Command Injection (HIGH): The skill executes a Grep command using {vault} and {folder} parameters without apparent sanitization. A malicious user or agent could inject shell commands via these variables to achieve arbitrary code execution.
Indirect Prompt Injection (HIGH): 1. Ingestion points: Reads all .md files in the vault via Grep (SKILL.md). 2. Boundary markers: Absent; link targets are parsed directly from file content without delimiters. 3. Capability inventory: Executes shell commands (Grep) and generates actionable recommendations based on note content. 4. Sanitization: None. Risk: Content within vault files could be crafted to manipulate the agent's reasoning or the resulting report recommendations, potentially leading to destructive actions if the agent has write access.
Data Exposure (LOW): The skill accesses all file content within the specified directory. While functionally necessary for link checking, it exposes the entire private knowledge base to the agent's context.

link-check