Add HOL Guard runtime safety skill

[kantorcodes]

Summary

Adds a repository-native hol-guard security skill that directly installs and invokes the HOL Guard runtime before mutation-bearing work on supported local AI coding harnesses.

• adds hol-guard/SKILL.md with fail-closed Guard-owned detection, bootstrap/install, dry-run, doctor, protected launch, approvals, receipts, and status verification
• pins HOL Guard 3.0.12 for reproducible optional setup: https://github.com/hashgraph-online/hol-guard/releases/tag/v3.0.12
• adds the skill to Hacktron's Claude Code plugin marketplace so the official registry can distribute it
• keeps native harness authentication/permissions/sandboxing authoritative and keeps Guard Cloud opt-in
• adds no generic framework primitive and does not claim hosted/server-side interception

Validation

• exact upstream HOL Guard issue/PR dedupe was zero immediately before opening this PR
• .claude-plugin/marketplace.json preserves the existing marketplace shape and registers ./hol-guard
• the skill follows the repository's documented Agent Skills frontmatter structure and security category
• the documented hacktron skills enable hol-guard runtime check was not claimed: the Hacktron CLI is not available in this execution environment, so this PR intentionally does not fabricate an interactive install attestation
• the 3.0.12 refresh updates the pinned HOL Guard release and removes evergreen current-stable wording; no new interactive runtime validation is claimed for that refresh

All skill content is reviewable text; no generated binaries or hidden setup scripts are included.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b1cd9f497c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".