Silent Skill Failures Expose Hidden Gaps in AI Agent Portability Tools
A developer migrating six Claude Code skills to OpenCode found that all files loaded without errors, yet one skill silently failed due to unsupported frontmatter fields like model pinning and dependency declarations. An audit revealed that Claude Code supports over 15 frontmatter fields while OpenCode recognizes only five, meaning harness-specific configurations are dropped without any warning. Five of the six skills ported cleanly only because they happened to use the minimal portable subset of name, description, and markdown body. Popular conversion tools such as farmage/opencode-skills and crosstrain move skill files between platforms but do not flag what functionality breaks in the process. The developer published an audit framework and a five-minute compatibility test to help users identify which skills transfer safely and which require a rebuild.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in