Skill evaluation
Tools that check whether an Agent Skill is loaded when it should be and whether it improves results over running without it. coder-eval, listed under Agent and harness benchmarks, also checks whether a given skill was loaded and runs A/B comparisons with and without skills.
- skill-creator - Anthropic’s skill for building Agent Skills, whose scripts run
claude -pon should-trigger and should-not-trigger queries to measure how often a skill’s description makes Claude load it, and aggregate with-skill and baseline runs into pass-rate, time, and token deltas; Apache-2.0 per the skill’s LICENSE.txt.