The Agent Skills Directory

Unverifiable Dependencies (HIGH): Under the 'Library-First Approach' section, the skill mandates that the agent 'ALWAYS search for existing solutions before writing custom code' and specifically 'Check npm for existing libraries.' This directive instructs the agent to fetch and potentially integrate external, untrusted code from public registries without specifying verification or version pinning requirements. This behavior creates a significant surface for supply chain attacks where the agent might inadvertently suggest or install compromised packages.
Indirect Prompt Injection (MEDIUM): The skill is designed to process user requests for 'writing code, design architecture, analyze code.' Because it lacks specific boundary markers or instructions to ignore embedded commands in the code it analyzes, there is a risk that malicious comments or documentation in the analyzed code could influence the agent's behavior (Category 8).
Command Execution (MEDIUM): The instructions to 'Use libraries instead of writing your own' and the mentions of specific ecosystems like npm implicitly require the agent to generate and potentially execute package installation commands (e.g., npm install). If the agent operates with high privileges, this can lead to the execution of malicious install scripts from untrusted packages.

software-architecture