The Agent Skills Directory

[NO_CODE]: The skill is entirely composed of Markdown methodology documents and YAML configuration. It lacks any executable scripts, binaries, or automated shell commands, which inherently limits the security risk profile.
[SAFE]: No evidence of data exfiltration, credential harvesting, or sensitive file access was found. All external URLs identified in the benchmark cards are plain-text links to public social media references for comparative analysis.
[PROMPT_INJECTION]: The instructions establish a rigorous evaluation framework for the AI agent and do not include patterns intended to bypass safety guardrails, extract system prompts, or override core instructions.
[SAFE]: The skill lacks the capabilities required for persistence, privilege escalation, or dynamic code execution. It functions purely as a qualitative analysis tool based on the provided internal knowledge base.

bmz-preflight