Skill Routing Benchmark

测试多个 Skill 描述是否会路由冲突，并生成正例、反例与负向触发语句。；use for skills, routing, benchmark workflows；do not use for 只给模糊建议, 忽略高度相近的 skill.

This skill appears coherent and low-risk: it runs a local Python script to analyze skill descriptions and produce structured reports. Before running: (1) review scripts/run.py yourself (it is included and readable); (2) avoid pointing the tool at directories or files that contain secrets or sensitive system configs — it will read and sample text files; (3) use the --dry-run and/or run in a sandboxed environment the first time; (4) verify outputs before copying or using them in any automated publish/workflow. There are no required credentials or network downloads in the bundle. If you need to be extra cautious, inspect the repo locally and run the script on a small, non-sensitive sample input first.

Type: OpenClaw Skill Name: skill-routing-benchmark Version: 1.0.0 The skill bundle is a legitimate utility designed to audit and benchmark AI skill routing and security patterns. The core logic in scripts/run.py processes local files to generate structured Markdown reports and includes a 'pattern_audit' mode that identifies (and partially masks) sensitive strings like API keys or dangerous shell commands. The behavior is entirely consistent with the stated purpose of skill quality assurance, and the code lacks network access, obfuscation, or unauthorized execution capabilities.