Superpowers vs Agent Skills

A label partly reflects how much of each tool is visible and scannable, not only how risky it is. A tool without a lockfile cannot get a dependency check.

• Superpowers: code and instruction-file checks found nothing, but there is no runtime lockfile or manifest, so dependencies were not checked
• Agent Skills: no issues found

Superpowers scores 9 points higher in Week 45: Superpowers 62, Agent Skills 53 out of 100. The score combines maintenance, adoption and docs (percentiles of public GitHub signals): Superpowers 22, 98 and 69; Agent Skills 12, 84 and 69. Community (20% weight) is not measured yet. It is not a hands-on test, so a higher score does not say which tool suits your project.

Safety labels from the automated scan of 2 Nov 2026: Superpowers: No dependencies scanned, code checks clean, no lockfile to check; Agent Skills: No issues found, no issues found in this week’s scan. “No dependencies scanned” means the code checks were clean but there is no lockfile, so dependencies could not be checked; it is not a clean bill. It is an automated scan, not a manual audit.

By GitHub stars: Superpowers 310,769, Agent Skills 186,457. Adoption score (percentile among all 52 tracked tools): Superpowers 98, Agent Skills 84. Stars are one public signal and do not measure real-world use.

Superpowers: MIT; Agent Skills: Apache-2.0 + proprietary. Read each licence text before commercial use.

Pick Superpowers if recent activity and releases matter most to you (maintenance 22); you want the more widely adopted option (adoption 98); you need a permissive open-source licence (MIT). Pick Agent Skills if you want the tool whose automated scan found no issues. This comes from measured public data and licences only, not from hands-on testing.