feat: Add native support for Hermes Agent skills format

[XiaotianTang3]

What does this PR do?

Adds native support for Hermes Agent, a powerful terminal/orchestration framework.

Hermes uses a strict SKILL.md format with YAML frontmatter. To avoid hallucinations (where the agent might try to use Claude Code slash commands like /review), this PR intelligently parses the persona markdown and wraps it into Hermes-compliant skills with an explicit SYSTEM NOTE instructing it to use native Hermes tools (terminal, read_file, etc.) instead.

Changes:

• Added convert_hermes() in scripts/convert.sh to package agents into SKILL.md format.
• Added install_hermes() in scripts/install.sh to route skills to ~/.hermes/skills/agency-agents/.
• Auto-detects Hermes via command -v hermes or ~/.hermes dir.
• Updated README.md to reflect Hermes support.

Agent Information (if adding/modifying an agent)

N/A - Core integration feature

Checklist

• Follows the agent template structure from CONTRIBUTING.md
• Includes YAML frontmatter with name, description, color
• Has concrete code/template examples (for new agents)
• Tested in real scenarios
• Proofread and formatted correctly

[msitarzewski]

Thanks for tackling Hermes! 🙏

Hermes support has now landed via #614, which took a lazy router-plugin approach: it installs a single agency-agents-router plugin and loads agents on demand through router tools, rather than front-loading the full 232-skill roster into the prompt (that design saves a big chunk of context — ~13k tokens). Because that approach was merged, I'm closing this one as superseded — no reflection on the work here, there were just several parallel Hermes PRs and we consolidated on the plugin route.

If you think the router approach is missing something your implementation had, or the design needs tweaking, let's iterate in a Discussion (Ideas category) rather than a parallel PR — happy to fold improvements into the merged integration. Thanks again!