Observed arrival · 2026-11-22
Skill Atlas: a field guide to agent skills
A public catalog for examining Claude Code Agent Skills, their source files, workflows, and any recorded evaluation evidence.
- For
- Developers evaluating Claude Code Agent Skills
- Worth noticing
- It separates fixed-commit source inspection from controlled-run measurements, and leaves run results absent when no real evaluation was recorded.
Field notes
The sample dossiers name their source collection and summarize intended workflows, then report structure counts and package size: each of the three shown has six files and no scripts, with sizes from 20.4 to 59.6 KB. The method page distinguishes a deterministic reading of a fixed commit from a controlled run; measured outcomes, token use, and duration are included only when a real evaluation has been recorded. The site explicitly says static analysis is not a security guarantee.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue