Add public-safe PM Agent OS skills

[whisperbot]

Summary

• Add public-safe PM Agent OS skill framework under skills/
• Add sanitization notes for public sync boundaries
• Update README with skill entry points
• Ignore internal working materials, generated exports, build artifacts, and local secrets

Sanitization

• Removed internal document URLs, internal domains, model gateway endpoint, team-specific names, experiment/dashboard references, and real API keys
• Replaced private knowledge sources with placeholders such as <TEAM_CONTEXT_INDEX_URL>, <MODULE_ONE_PAGE_URL>, and <PRD_TEMPLATE_URL>
• Kept methods and workflows only; dynamic knowledge remains in private One Pages

Validation

• Ran staged sensitive-term scan for internal domains/names and key patterns
• Ran Python compile check for the multimodal helper
• Verified skill frontmatter names match public skill folders