Empirical agent skills for AI coding assistants — Claude Code, Cursor, Codex, Cline, Gemini CLI, and any other tool implementing the agentskills.io spec.
Every skill is derived from hands-on integration work, not just docs.
Every gotcha has a reproducible repro path documented in the skill itself,
plus a minimal runnable example under examples/<skill>/.
# all skills
npx skills add ophelios-studio/skills
# one specific skill
npx skills add ophelios-studio/skills --skill 0g
npx skills add ophelios-studio/skills --skill axl
npx skills add ophelios-studio/skills --skill axl-pubsub
npx skills add ophelios-studio/skills --skill leaf
npx skills add ophelios-studio/skills --skill zephyrusThe CLI installs to ~/.agents/skills/<name>/ and symlinks into your
agent's skill directory (e.g. ~/.claude/skills/).
| Skill | What it covers |
|---|---|
0g |
0G Chain (Galileo, 16602), Storage, Compute — empirical patterns |
axl |
Gensyn AXL HTTP API — protocol primitives, library-agnostic |
axl-pubsub |
The axl-pubsub gossip library on top of AXL — when and how |
leaf |
Zephyrus Leaf static-site generator |
zephyrus |
Zephyrus PHP framework |
skills/<name>/SKILL.md # the skill itself (frontmatter + body)
examples/<name>/ # minimal runnable scripts a dev can verify
Skills written from docs alone go stale fast. These skills capture what actually happens when you run real code against real testnets and real infrastructure — empirical reality first, official documentation second.
Every gotcha in every skill is reproducible: cite the file:line of the production code that taught us, link the live transaction or test run that proves it.
MIT.