Adopt
OpenAI's official skills catalog — proof the SKILL.md format is a true cross-vendor standard.
Why It Matters
github.com/openai/skills uses the same SKILL.md format originated by Anthropic. Compatible with Codex CLI and ChatGPT. When your biggest competitor adopts your format, that format wins. This repo validates that SKILL.md is not an Anthropic proprietary play but a genuine industry standard.
Strengths
- Same SKILL.md format means skills are portable across Claude Code, Codex, and more
- First-party OpenAI backing gives confidence in long-term support
- Validates cross-vendor interoperability — write once, run on any compatible agent
Limitations
- Catalog is smaller than Anthropic's; OpenAI entered the skills game later
- Some skills are optimized for Codex/ChatGPT behavior and may need tweaking elsewhere
- Like the Anthropic repo, no built-in security scanning at the format level
Risks
- "Cross-vendor standard" obscures the reality: OpenAI adopted the format strategically, not altruistically — it's cheaper to embrace than to compete on format
- Skills optimized for Codex's behavior patterns can produce subtly wrong results on Claude Code or Gemini CLI without obvious errors
- The catalog is growing slowly; OpenAI's primary investment is in the GPT Store and Codex plugins, not SKILL.md
- Same security concerns as Anthropic Skills: no sandboxing, no permissions, no trust boundaries — markdown all the way down