coding-agent-security-benchmark

Produces publishable markdown benchmark reports from eval results, modeled on the structure and tone of frontier AI lab reports (Anthropic, Cursor, FrontierSWE, BaxBench). Use this when the user wants to write up benchmark findings, turn eval results into a blog post or technical report, publish a model/agent comparison, announce a new eval, or share benchmark data publicly. Triggers on phrases like "write a benchmark report", "turn these results into a post", "publish our evals", "draft a writeup of the benchmark", "create a model comparison report", "generate a benchmark blog post". Use this skill even if the user just says "write this up" or "make a report" in the context of benchmark/eval results. Do NOT use for internal changelog entries, release notes, PR descriptions, or academic paper drafts.

Creates high-quality agent skills by running the official Claude skill creator plugin and then applying a structured set of enhancement phases covering use-case design, frontmatter completeness, body structure, writing quality, and progressive disclosure architecture. Use this skill when the user wants to create a new skill, improve an existing SKILL.md, or asks "how do I write a good skill", "help me make a skill for X", "review my SKILL.md", or "turn this workflow into a skill". Trigger even if the user just says "let's make a skill" or describes a workflow they want to capture.