When you're wrangling a repo full of SKILL.md files and need them to actually conform to the Agent Skills spec, this is your audit and repair loop. It validates frontmatter, checks trigger descriptions for weakness, catches catalog drift between live folders and your discovery surfaces, and handles the canonicalization dance when two skills compete for the same job. The working modes are unusually precise: you pick validate, create, canonicalize duplicates, structural sync, or derived surface sync, then run the truthful validator command for your current directory. It treats natural language in descriptions as sacred and validator bugs as fixable, which is the right stance. If you've ever shipped a skill rename only to find stale paths in three different README files, the residue scan checklist will feel familiar.
npx -y skills add akillness/oh-my-skills --skill skill-standardization --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills