Add OpenCode runtime compatibility

Context

OpenCode (https://github.com/anomalyco/opencode) is a fast-growing AI coding agent that natively discovers skill files from .claude/skills/ and ~/.config/opencode/skills/. Its skill frontmatter schema is:

name (required)
description (required)
license (optional)
compatibility (optional) — recommended for discovery optimization
metadata (optional, string-to-string map) — arbitrary key-value for custom fields
(all other fields are ignored)

This repo's skills already work in OpenCode when placed in .claude/skills/ (OpenCode discovers that location). The skills' deliberate portability (no $ARGUMENTS, plain ## Input sections, multi-runtime intent) makes them an excellent fit.

Proposed scope

1. Add compatibility: opencode to all 55 SKILL.md frontmatter — inert field, signals OpenCode discovery.
2. Nest rich fields (intent, type, theme, best_for, scenarios, estimated_time) under metadata: — aligns with OpenCode's schema; preserves all semantics; OpenCode agents can still read them even though the skill tool only indexes name and description.
3. Keep argument-hint as top-level (Claude Code autocomplete relies on it; OpenCode ignores it harmlessly).
4. Add ~3 lines to README: copy skills to ~/.config/opencode/skills/ or .opencode/skills/.
5. Update scripts/check-skill-metadata.py to recognize compatibility and metadata as valid fields.

Why this fits the repo's philosophy

This repo explicitly values portability across runtimes (the CONTRIBUTING.md rejects $ARGUMENTS for this reason). Adding OpenCode as an explicit runtime extends that portability promise without introducing runtime-specific syntax.

Non-goals

• No OpenCode-specific invocation syntax (no !command, no runtime macros)
• No changes to skill body content
• No Claude Code marketplace changes

Ready to PR if maintainer is open to it.