The Agent Skills Directory

[SAFE]: The skill consists of instructional content for securing AI workflows using the 'Nika' toolset. It defines defensive patterns for agent operations, including explicit budget ceilings, least-privilege authorization boundaries (permits), and mandatory masking of environment-injected credentials. No malicious code, prompt injections, or unauthorized network operations were detected.
[INDIRECT_PROMPT_INJECTION]: The documentation identifies a vulnerability surface involving Model Context Protocol (MCP) servers (ingestion point) and provides instructions for tool pinning (boundary marker) to prevent unauthorized tool execution (capability). It requires manual human approval (sanitization) to accept tool definition changes, effectively managing the risk of external data poisoning.

nika-operating