A user conducted an experiment to evaluate the effectiveness of nine popular "Claude Code skills," which are essentially text additions to Claude's context. Using a placebo control method similar to drug trials, the user tested these skills against neutral instructions of the same token length. The results showed that only two skills, "ponytail" and "agent-skills," performed better than their placebo, while one skill, "planning-with-files," performed worse. The other six skills showed no significant improvement over the placebo, and none of the skills made Claude Code measurably cheaper to run. AI
IMPACT User-generated analysis suggests that many third-party "skills" for Claude Code may not offer significant benefits over basic instructions.
RANK_REASON User-led analysis of a product's features, not a primary release or research paper.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →