Improve this SKILL.md so Claude exhibits the expected behavior.

## Current SKILL.md

${current_skill}

## Failures

These prompts did NOT produce the expected behavior:

${failures_yaml}

## Your Task

Revise the SKILL.md to fix these failures. Common issues:
- Description not specific enough about WHEN to trigger
- Instructions not explicit enough (Claude defaults to asking permission)
- Missing "do NOT ask" or "IMMEDIATELY" language for automatic behaviors

IMPORTANT CONSTRAINTS:
- Keep the SKILL.md under 500 lines (Anthropic's recommended limit)
- Be concise - shorter is better
- Replace verbose instructions with terse, direct ones
- Do NOT keep adding more text - rewrite to be minimal
${tip_section}
Return ONLY the improved SKILL.md content (no explanation, no code fences).
