Add HOL Guard safety skill for agent workflows

[kantorcodes]

Summary

Adds hol-guard as an optional safety skill in the official Oxylabs Agent Skills bundle.

The skill installs and invokes HOL Guard at the supported local AI harness boundary before credential-bearing, cost-bearing, or state-changing Oxylabs workflows, then routes the actual work to the existing Oxylabs product skill (proxies, web-unblocker, web-scraper-api, headless-browser, or video-data).

It explicitly does not claim that HOL Guard runs inside Oxylabs services or replaces Oxylabs authentication, limits, billing controls, legality checks, or request validation.

Distribution

This adds ./skills/hol-guard to the existing Claude Code plugin manifest and README catalogue. If merged, the existing installation paths distribute the skill with the Oxylabs bundle:

/plugin marketplace add oxylabs/agent-skills
/plugin install oxylabs@oxylabs-agent-skills

and:

npx skills add https://github.com/oxylabs/agent-skills.git

Validation

• branch is 0 behind / 2 commits ahead of current main
• .claude-plugin/plugin.json and marketplace JSON parse successfully
• git diff --check passes
• exact upstream issue/PR/code searches for HOL Guard / hol-guard were clean immediately before submission
• HOL Guard commands were checked against its maintained canonical Agent Skill
• no live Oxylabs API/browser mutation was run or claimed

Affiliation: I maintain HOL Guard / Hashgraph Online. AI assistance was used to prepare the focused contribution; the Oxylabs/HOL Guard boundary claims were source-checked before submission.