# Orditra integration for OpenCode

Follow the shared Orditra core policy. OpenCode discovers portable skills in
`~/.agents/skills`; do not maintain a third duplicate skill tree.

Use the context-mode plugin for large output and continuity, Serena MCP for
semantic code operations, ast-grep for structural syntax, and `rg` for text.

Check project bootstrap prerequisites before tracker- or domain-dependent
workflow skills.

Keep remote MCP endpoints explicit in OpenCode JSON and never write tokens into
the shared repository. Project-only capabilities belong in repository config,
not the global OpenCode tool list.