Skill Sharer

Publishes a local skill to /Enterprise-Crew-skills on GitHub.

What it does

1. Copies the skill into a sanitized folder
2. Strips personal information, secrets, IPs, paths, and credentials
3. Generates a standalone README for the skill
4. Updates the repo's root README with the new skill entry
5. Commits and pushes

Usage

# Share a skill (interactive — reviews sanitization before pushing)
~/agent-workspace/skills/skill-sharer/scripts/share-skill.sh <skill-folder-path> [--description "Short description"]

Examples

# Share session-cleaner
~/agent-workspace/skills/skill-sharer/scripts/share-skill.sh ~/agent-workspace/scripts/session-cleaner/ --description "Converts session JSONL to clean markdown"

# Share a skill from the skills directory
~/agent-workspace/skills/skill-sharer/scripts/share-skill.sh ~/agent-workspace/skills/weather/ --description "Get weather forecasts with no API key"

Sanitization rules

The script strips:

• IP addresses — Tailscale IPs, public IPs, local IPs (replaced with <REDACTED_IP>)
• Paths with usernames — /home/user/, /home/user/ → generic paths
• API keys and tokens — anything matching key/token/secret patterns
• Email addresses — real emails replaced with user@example.com
• SSH connection strings — ssh user@host → ssh user@<your-host>
• Server URLs with real hosts — internal URLs replaced with placeholders
• Secret file references — ~/agent-workspace/secrets/* → <YOUR_SECRET_FILE>
• Tailscale hostnames — machine names replaced
• Environment variable values — actual values stripped, variable names kept

Agent workflow

When Henry asks to share a skill:

1. Identify the skill folder path
2. Run share-skill.sh <path> --description "..."
3. Review the sanitized output (script pauses for review)
4. Confirm to push
5. Report the GitHub URL to Henry

Repo structure

Enterprise-Crew-skills/
├── README.md              ← Root index (auto-updated)
├── session-cleaner/
│   ├── README.md
│   └── ...
├── new-skill/
│   ├── README.md
│   └── ...