Hermes Skill Inventory Audit Report — May 2026
Reviewed: June 4, 2026
Generated: May 26, 2026 | Total Skills: 279
This report audits the Hermes agent skill inventory, identifying active vs. archived skills, overlaps, gaps, and actionable optimization recommendations.
Executive Summary
| Metric | Count |
|---|---|
| Total Skills | 279 |
| Active Skills | 160 |
| Archived Skills | 59 |
| Skills with Categories | 130 |
| Skills with Placeholder Description | 1 |
Category Breakdown
| Category | Count | Notes |
|---|---|---|
| devops | 38 | Largest category — infrastructure & deployment |
| mlops (all subcats) | 32 | ML training, inference, evaluation, models |
| content | 10 | Writing, design, publishing workflows |
| creative | 17 | Art, diagrams, video, music generation |
| autonomous-ai-agents | 6 | Coding agents, memory systems |
| software-development | 11 | Coding workflows, debugging, testing |
| productivity | 9 | Documents, email, calendar, meetings |
| integration | 8 | M365, Telegram, Ollama, Exchange |
| wordpress | 4 | WP management, publishing, troubleshooting |
| media | 8 | YouTube, Spotify, GIFs, audio analysis |
| research | 6 | arXiv, blogs, cost analysis, markets |
| github | 5 | Repos, PRs, issues, auth |
| content-strategy | 1 | Content catalog methodology |
| copywriting | 1 | Automated service messaging |
| python | 2 | Package management, import fixes |
| .archived | 59 | Retired but preserved for reference |
Key Findings
1. Duplicate/Overlapping Skills
- mlops-inference (archived) overlaps with serving-llms-vllm, gguf-quantization, llama-cpp, and obliteratus (all active/mlops). The archived skill should be removed or merged.
- inference (archived) is fully superseded by the mlops/inference category skills.
- state-cleanup (archived) — functionality is built into the masterdash-execution-pattern skill. Safe to remove.
- wave-execution (archived) — superseded by masterdash-execution-pattern. Safe to remove.
- auto-backlog-grooming (archived) — patterns absorbed into masterdash-execution-pattern.
- humanizer (content, REMOVED) — consolidated into creative/humanizer. The old entry is a redirect.
2. Skills Needing Description Updates
- pm_agent — has only „—“ as placeholder description. Should be updated with proper metadata or archived if no longer relevant.
- pm-agent-health-monitor — description starts with „[SILENT]“ which is a workaround artifact, not a proper description.
- humanizer (content) — marked as REMOVED, consolidated into creative/humanizer. Should be deleted from content/ category.
3. Operational Skills Assessment
The most critical operational skills for daily Hermes operation are well-maintained:
- masterdash + masterdash-execution-pattern — core execution engine ✅
- report-to-published-page — essential for publishing workflow ✅
- wordpress-publisher + wordpress — content publishing ✅
- hermes-agent — self-configuration ✅
- software-development cluster — coding workflows ✅
4. Recommended Actions
- Archive cleanup: Remove 5 superseded archived skills (mlops-inference, inference, state-cleanup, wave-execution, auto-backlog-grooming).
- Fix pm_agent: Update description or archive if no longer relevant.
- Merge devops/mlops: Consider merging devops and mlops categories into a single „infrastructure“ category since both handle deployment.
- Add testing skill: The automated-testing skill is archived but there’s no active replacement for test automation workflows.
- Review trading skills: 6 trading-related skills are archived. If trading is not an active use case, remove entirely. If active, update and un-archive.
Health Score
Overall: 8.2/10 — The skill inventory is well-organized with clear categorization. Main issues are superseded archived skills creating noise and two skills with missing descriptions. The core operational skills are comprehensive and up-to-date.