feat: add HOL Guard runtime safety skill

[kantorcodes]

Summary

• add a hol-guard Agent Skill to SkillX's maintained Claude Code marketplace
• guide users through isolated HOL Guard installation and the detected-harness bootstrap / install / run --dry-run / run / doctor / status flow
• include approval, receipt, evidence, and separate plugin-scanner package-verification guidance without treating scanner results as a safety guarantee

Direct install outcome

If merged, SkillX users can install the maintained entry through:

/plugin marketplace add nextlevelbuilder/skillx
/plugin install hol-guard@skillx-marketplace

The installed skill then directly invokes HOL Guard for supported local AI harnesses rather than only linking to documentation.

Validation

• .claude-plugin/marketplace.json parses as valid JSON
• commands and supported harness names were checked against the current HOL Guard Agent Skill source
• the skill uses direct CLI version probes instead of POSIX-only command -v checks
• full pnpm test / pnpm lint was not run in this execution environment; this contribution changes marketplace JSON and Markdown only

Source project: https://github.com/hashgraph-online/hol-guard
License: Apache-2.0