feat: add HOL Guard Vercel command safety skill

Summary

Adds hol-guard-vercel, a focused Agent Skill for applying HOL Guard before production-impacting Vercel CLI operations.

• adds skills/hol-guard-vercel/SKILL.md
• groups the skill under Vercel in skills.sh.json
• documents the standard npx skills add vercel-labs/agent-skills --skill hol-guard-vercel install path

Why

HOL Guard release/3.0 includes command.platform.vercel coverage for removing deployments/projects and production deploy, promote, and rollback operations, while preserving documented inspection/help counterparts. The skill uses that existing command-safety surface rather than asking this repo for new framework behavior.

The workflow prefers a Guard-protected supported coding-agent harness for real review/approval handling. hol-guard command test is presented only as a side-effect-free preflight and is explicitly not described as a final runtime policy or approval decision.

Scope and safety

• no Vercel API or runtime changes
• no claim of native Vercel interception
• review / block / unknown / error paths are fail-closed
• no Guard Cloud dependency
• no production Vercel mutation is performed by the skill itself

Verification

• directory name matches frontmatter name
• skill is under the repository's 500-line guidance
• skills.sh.json remains valid JSON with the skill in the Vercel grouping
• Vercel operation coverage checked against HOL Guard's current release/3.0 coverage docs

Affiliation: I maintain HOL Guard / Hashgraph Online. AI assistance was used to prepare this contribution; the commands and scope were checked against the current source/docs before submission.