!ACTIVE=$(cat personas/_active.md 2>/dev/null || echo "default"); cat "personas/$ACTIVE/persona.md" 2>/dev/null && cat "personas/$ACTIVE/nuance.md" 2>/dev/null || echo "No persona. Run /interview first."

Export Skill Protocol

$ARGUMENTS:

• target — claude, agents, or both

Parse target first:

• claude → write only personas/{active}/generated/CLAUDE.generated.md
• agents → write only personas/{active}/generated/AGENTS.generated.md
• both or empty → write both generated files

Export to CLAUDE.md template (target=claude or both)

Build personas/{active}/generated/CLAUDE.generated.md from persona + nuance sources:

• Work philosophy section: infer priority, approach, and risk stance from persona.md
• Decision-making priorities: infer ordering + energy bias from persona.md
• Tone rules: map nuance.md voice into register rules and avoid-list
• Context loading protocol: derive startup context from persona.md core tags
• Anti-patterns to avoid: convert persona.md anti field into explicit do-not rules
• Add a header note that this file was generated and should be reviewed before copying into a project root.

Export to AGENTS.md template (target=agents or both)

Build Codex/OpenCode-compatible agent definitions:

• Include agent name, description, and invocation method
• Ground role split in persona domain expertise from persona.md core tags
• Keep entries portable across Codex/OpenCode style AGENTS.md
• Write the result to personas/{active}/generated/AGENTS.generated.md and add a generated-file header note.

Output

Write generated files to personas/{active}/generated/:

• CLAUDE.generated.md
• AGENTS.generated.md
• Create the directory first if it does not exist.
• Re-read the saved file(s) and return the confirmed generated path(s).

Note: This persona is now portable. Drop it into any project.