| name | orchestration/skill-navigator |
| trigger-words | ["orchestration/skill","navigator"] |
| description | Quick-reference navigator for all 169 active Hermes skills. Groups skills by domain-cluster (multi-agent, greyhack, mlops, creative, etc), tells you which skill to load for which task, flags clusters with overlap or gaps. Built from a real scan on 2026-07-02. Companion to the skill_view function — load this FIRST to decide which skill to load next.
Trigger phrases: orchestration/skill, navigator.
Converted from Hermes-Skill (Yuno's mastery library).
|
🔥 Top-10 Most-Used Clusters (Decision Shortcuts)
⚡ NEW-USER INSTRUCTION: If you (Yuno or other agent) just loaded this skill, you probably need one of the clusters below. The cheat-sheet table maps user intents to primary + companion skills. If unsure between two skills, load the narrower one first (cheatsheet > domain > meta).
When a user request matches one of these patterns, load the bolded primary skill first:
| User Intent | Primary Skill | Companion Skills |
|---|
| Spawn parallel research/audit agents | multi-agent-pitfalls-cheatsheet | multi-agent-orchestration, multi-agent-work |
| Set up or debug Hermes itself | hermes-agent | hermes-maintenance, messaging-gateway-setup, hermes-admin |
| Configure Telegram/Discord gateway | messaging-gateway-setup | hermes-gateway, hermes-copilot-setup |
| Git workflow (commit, PR, branch, issues) | github-pr-workflow | github-issues, github-code-review, github-auth |
| Write/refactor/debug code | subagent-driven-development | plan, critic-gate, simplify-code, test-driven-development |
| GreyHack scripting | greyhack | greyhack-greyscript, greyhack-sandbox, greyhack-hermes-api |
| ML/Ollama/Local AI setup | ollama-local-hosting | mlops-suite, llama-cpp, local-ml-hosting |
| Build creative HTML/SVG/design | creative-suite | claude-design, popular-web-designs, html-artifact |
| System maintenance / cleanup | linux-system-maintenance | host-security-audit, local-ai-security-hygiene |
| Document writing / conversion | system-documentation | epub-export, pdf, obsidian, notion |
📊 Skill-Clusters (Domain-Grouped)
🐝 Cluster 1: Multi-Agent & Orchestration (8 skills)
MULTI-AGENT CLUSTER — when spawning parallel work
├── multi-agent-pitfalls-cheatsheet ← LOAD FIRST (trap-watchlist)
├── multi-agent-orchestration ← pattern + 5 phases
├── multi-agent-work ← end-to-end slash command
├── research-orchestration ← research-heavy plans
├── kanban-orchestrator / worker ← board-based long-running
├── hermes-agent ← general Hermes orchestration
├── the-dmz-transfer ← transfer from external source
└── depps-orchestration ← low-trust workers
When to load what:
- Just spawning subagents? →
multi-agent-pitfalls-cheatsheet
- Full research project? →
multi-agent-orchestration + cheatsheet
- One-shot research? →
/multi-agent-work slash command
- Multi-day kanban board? →
kanban-orchestrator
🛠️ Cluster 2: Software Development (26 skills — LARGEST!)
CODE/DESIGN/TEST CLUSTER
├── Code-writing & patterns
│ ├── subagent-driven-development ← plan → delegate → execute
│ ├── plan ← write implementation plan
│ ├── critic-gate ← output quality gate
│ ├── simplify-code ← parallel cleanup agents
│ ├── refactor / code-mod ← safe refactors
│ └── vanilla-js-tdz-helper-first ← JS helpers
│
├── Code-review & QA
│ ├── requesting-code-review ← pre-commit review
│ ├── bash-script-audit ← bash safety scanner
│ ├── security-code-checker ← LLM-generated-code risks
│ ├── output-validator ← syntax pre-flight
│ └── systematic-debugging ← 4-phase root cause
│
├── TDD & Testing
│ ├── test-driven-development ← RED-GREEN-REFACTOR
│ ├── python-debugpy ← remote debugger
│ ├── node-inspect-debugger ← Node inspector
│ └── rag-pipeline-python ← RAG TDD example
│
├── Language-specific
│ ├── ki-murks-verhindern ← AI workflow gates
│ └── hermes-agent-skill-authoring ← writing Hermes SKILL.md
│
├── Spike/Experiment
│ └── spike ← throwaway experiments
│
└── Other
├── coding-skills-audit
└── multi-agent-research ← archived pattern source
🎨 Cluster 3: Creative & Design (23 skills)
CREATIVE CLUSTER
├── Suite / Orchestrator
│ └── creative-suite ← full UI-factory chain
│
├── UI Components
│ ├── ui-component-library ← buttons, inputs, cards
│ ├── ui-color-system ← accessible palettes
│ ├── ui-design-system ← full design tokens
│ ├── ui-factory ← composite UI workflows
│ └── ui-dashboard ← data-schema → KPI dashboard
│
├── Web design & artifacts
│ ├── claude-design ← one-off HTML artifacts
│ ├── html-artifact ← self-contained HTML
│ ├── popular-web-designs ← 54 real design systems
│ ├── web-design-guidelines ← Vercel Web review
│ ├── pretext ← @chenglou/p pretext browser
│ ├── design-md ← Google DESIGN.md spec
│ └── sketch ← 2-3 variant mockups
│
├── Diagrams & Visual
│ ├── architecture-diagram ← dark-themed SVG
│ ├── ascii-art ← pyfiglet, cowsay
│ ├── ascii-video ← video → ASCII
│ ├── excalidraw ← hand-drawn diagrams
│ ├── manim-video ← math animations
│ └── pixel-art ← NES/GB/PICO-8
│
└── Other
└── humanizer ← strip AI-isms from text
⚙️ Cluster 4: DevOps & Maintenance (18 skills)
DEVOPS CLUSTER
├── Hermes maintenance
│ ├── hermes-maintenance ← update side-effects
│ ├── hermes-admin ← CLI/config troubleshooting
│ ├── hermes-v7-sse ← V7 SSE package
│ ├── hermes-v7-sse-server ← V7 SSE server
│ └── hermes-mcp-integration ← plugin-registry pattern
│
├── Linux system
│ ├── linux-system ← system troubleshooting
│ ├── linux-system-maintenance ← safe cleanup CLI
│ ├── linux-display-setup ← EDID / modelines
│ ├── nvidia-laptop-gaming-tuning ← Optimus tuning
│ ├── nvidia-suspend-resume ← GPU sleep/wake
│ └── docker-install-ubuntu ← Docker on Ubuntu
│
├── Security
│ ├── host-security-audit ← fwupd HSI / TPM / Secure
│ ├── local-ai-security-hygiene ← secure local-AI setup
│ ├── security-audit ← full security audit
│ ├── system-security-audit ← ports/permissions audit
│ └── bash-script-audit ← script safety
│
└── Messaging gateway
├── hermes-gateway ← gateway troubleshooting
└── messaging-gateway-setup ← platform setup
🎮 Cluster 5: Gaming (8 skills)
GAMING CLUSTER
├── GreyHack (the main cluster)
│ ├── greyhack ← meta-skill + pipeline
│ ├── greyhack-greyscript ← language reference
│ ├── greyhack-sandbox ← local Python toolkit
│ ├── greyhack-hermes-api ← in-game Hermes integration
│ ├── greyhack-greyscript-deep-audit
│ └── greyscript-compiler-debugging
│
├── Minecraft
│ ├── minecraft-modpack-server
│
├── Other
│ ├── pokemon-player
│ ├── teamspeak-server
│ └── game-library-management
🐙 Cluster 6: GitHub Workflow (7 skills)
GITHUB CLUSTER
├── github-pr-workflow ← branch → PR → merge
├── github-issues ← CRUD via gh CLI
├── github-code-review ← PR review + comments
├── github-auth ← HTTPS/SSH setup
├── github-repo-management ← clone/fork/remote
├── codebase-inspection ← pygount stats
└── github-workflow ← full gh CLI workflow
🤖 Cluster 7: MLOps (11 skills)
MLOPS CLUSTER
├── Inference & Serving
│ ├── serving-llms-vllm ← vLLM serving
│ ├── llama-cpp ← GGUF local inference
│ └── ollama-local-hosting ← Ollama setup/manage
│
├── Models
│ ├── huggingface-hub ← HF model search/download
│ ├── audiocraft-audio-generation ← MusicGen/AudioGen
│ └── segment-anything-model ← SAM zero-shot
│
├── Research & Evaluation
│ ├── dspy ← declarative LM programs
│ ├── evaluating-llms-harness ← lm-eval-harness
│ ├── llm-evaluation-troubleshooting
│ ├── weights-and-biases ← experiment tracking
│
└── Other
├── local-ml-hosting
└── obliteratus ← abliterate refusals
🎬 Cluster 8: Media (6 skills)
MEDIA CLUSTER
├── Audio
│ ├── spotify ← Spotify CLI
│ ├── heartmula ← song generation
│ ├── songsee ← audio spectrograms
│ └── audiocraft-audio-generation ← also in mlops
│
├── Video & Image
│ ├── gif-search ← Tenor GIFs
│ └── media-tools ← YouTube transcripts etc.
│
└── Other
└── youtube-content ← YT transcript → blog
🤖 Cluster 9: Autonomous AI Agents (6 skills)
AUTONOMOUS-AGENTS CLUSTER
├── claude-code ← delegate to Claude Code CLI
├── codex ← delegate to OpenAI Codex
├── opencode ← delegate to OpenCode
├── coding-agents ← general coding agent orchestrator
├── hermes-agent ← (also in orchestration)
└── the-dmz-transfer
📊 Cluster 10: Productivity (16 skills)
PRODUCTIVITY CLUSTER
├── Daily Operations
│ ├── daily-briefing ← system status briefing
│ ├── petdex ← animated mascots
│ └── teams-meeting-pipeline ← Teams summary
│
├── Documents
│ ├── epub-export ← MD/PDF → EPUB
│ ├── pdf ← PDF reading
│ ├── nano-pdf ← PDF editing
│ ├── ocr-and-documents ← extract text from scans
│ ├── powerpoint ← .pptx manipulation
│
├── External Services
│ ├── airtable ← REST API via curl
│ ├── google-workspace ← Gmail/Drive/Docs/Sheets
│ ├── linear ← issues/projects
│ ├── maps ← OSM geocoding
│ ├── notion ← Notion API
│
├── Note-Taking
│ ├── obsidian ← Obsidian vault
│ └── system-documentation ← MD doc tree
│
└── Email
└── himalaya ← IMAP/SMTP CLI
🚨 Edge-Case Singletons (1-skill categories)
These are alone in their category — load directly when relevant:
| Skill | Use When |
|---|
apple-notes / apple-reminders | macOS-only Apple ecosystem access |
arxiv | arXiv paper search |
architecture-diagram (root) | SVG cloud/infra diagrams |
ascii-video | Convert video/audio to ASCII |
blogwatcher | RSS/feed monitoring |
comfyui | ComfyUI local AI workflows |
computer-use | Desktop automation (background) |
context-mode | Context window optimization |
copilot-cli | GitHub Copilot CLI delegation |
data-science | Jupyter kernels |
debugging-hermes-tui-commands | Hermes TUI slash-command debugging |
dogfood | Exploratory QA of web apps |
excalidraw | Hand-drawn Excalidraw JSON |
firecrawl-web | Web scraping + screenshots |
find-skills | Discover/install new skills from skills.sh |
gif-search | Tenor GIF search |
google-oauth-setup | Google Workspace API OAuth setup |
ideation | Generate project ideas |
image-generate (root) | AI image generation |
kanban-codex-lane | Kanban + Codex CLI integration |
linux-display-setup | Custom EDID/modelines |
mcp-server-authoring | Build custom MCP servers |
mnemosyne-memory-provider | Mnemosyne native memory |
model-selector | Compare LLM models on Nous Portal |
multi-agent-orchestration | (also in orchestration cluster) |
native-mcp | Built-in MCP client |
notion | Notion API |
ollama-local-hosting | (also in mlops cluster) |
pdf-anthropic | PDF reading via Anthropic |
pokemon-player | Play Pokemon via headless emulator |
python-debugpy | Python remote debugging |
rag-pipeline-python | RAG pipeline (Tutor-style) |
red-teaming-godmode | LLM jailbreak testing |
research-orchestration | (in orchestration cluster) |
session-handoff | Model-switch handoff doc |
setup-guides-browser | Browser setup walkthroughs |
skill-install-workflow | skills.sh skill installer workflow |
songwriting-and-ai-music | Songwriting + Suno AI |
subagent-driven-development | (in software-development cluster) |
system-documentation | (in productivity cluster) |
telegram-clarification-prompt | Yuno's Telegram decision routing |
touchdesigner-mcp | TouchDesigner control |
voice-assistant-bots | Voice bot architecture |
web-archive-research | Common Crawl + WARC query |
web-design-guidelines | (in creative cluster) |
yuanbao | Yuanbao groups |
yuno-cleaner | Yuno Cleaner safe cleanup |
📐 Compliance Summary
Scan results (2026-07-02):
| Issue | Count | Severity | Action |
|---|
/home/bratan/ hardcoded paths | 21 files | LOW (cosmetic) | Optional: replace with ~/ |
| CLI drift (hermes skill/tools singular) | 0 | — | Clean |
| Malformed frontmatter | 0 | — | Clean |
| Read errors | 0 | — | Clean |
The 21 hardcoded-path files (all in skill examples, not in code paths):
- gaming/greyhack, gaming/greyhack-greyscript, gaming/greyhack-hermes-api
- software-development/hermes-agent-skill-authoring, software-development/bash-script-audit, software-development/subagent-driven-development
- devops/hermes-maintenance, devops/local-ai-security-hygiene
- research/web-archive-research, research/research-tools
- autonomous-ai-agents/claude-code, autonomous-ai-agents/codex
- productivity/daily-briefing, productivity/google-workspace
- mlops/huggingface-hub
- orchestration/multi-agent-orchestration
- And a few others
Recommendation: Not worth fixing — these are intentional examples showing actual user paths for pedagogical clarity. Skill consumers should mentally substitute ~.
🎯 Load-Order Heuristics (when in doubt)
- Trap-detector first →
multi-agent-pitfalls-cheatsheet for any subagent work
- Pattern-source second →
multi-agent-orchestration for the full pattern
- Domain-specific third → e.g.
greyhack for GreyHack tasks
- Specialty-fourth → e.g.
greyhack-greyscript for deep language reference
- Tool/script-specific fifth → e.g.
ci-build.sh for build pipeline questions
If you're not sure between two skills: load the one with the narrower scope first (cheatsheet > domain > meta).
📅 When to re-scan this Navigator
Re-run the compliance scan + re-build this cheatsheet when:
- New Hermes version adds/removes skills (check
hermes skills list)
- User adds a custom skill
- Major skill consolidation (like 2026-07-02 multi-agent-research → cheatsheet)
Re-scan from the reusable script
python3 ~/.hermes/skills/orchestration/skill-navigator/scripts/skill-scan.py
Quick-check (just counts)
python3 -c "import pathlib; print(sum(1 for f in pathlib.Path('/home/bratan/.hermes/skills').rglob('SKILL.md') if '.archive' not in f.parts and 'node_modules' not in str(f)), 'active skills')"
🔗 Cross-Domain Navigation (Domain ↔ Method Linking)
Pattern discovered 2026-07-02: Skills in domain clusters (gaming, devops, mlops) should cross-reference skills in method clusters (orchestration, software-development) to enable tool-discovery chains.
Why it matters
A user working on GreyHack (gaming/greyhack) who needs subagent support should not have to know that multi-agent-pitfalls-cheatsheet lives in orchestration/. The domain skill should surface the method skill via a "Related Skills" or "Cross-Cluster Navigation" section.
When to add cross-references
| Condition | Action |
|---|
| Domain skill describes a workflow that could use subagents | Add ref to multi-agent-pitfalls-cheatsheet + multi-agent-orchestration |
| Domain skill mentions parallel research/audit | Add ref to research-orchestration |
| Domain skill involves code review/QA | Add ref to requesting-code-review or systematic-debugging |
| Domain skill references a meta-navigator | Add ref to skill-navigator (this skill) |
| Method skill lists use-cases for multiple domains | Add refs to relevant domain skills in its "See Also" section |
Format template (add to target SKILL.md)
## 🧭 Related Skills (Cross-Cluster Navigation)
Skills that support this [domain] cluster but live elsewhere:
- **`skill-navigator`** (orchestration/) — Meta-Navigator. Load FIRST when deciding which skill applies.
- **`multi-agent-pitfalls-cheatsheet`** (orchestration/) — TRIGGER-WATCHLIST for `delegate_task` calls. Load BEFORE spawning subagents.
- **`multi-agent-orchestration`** (orchestration/) — The 3-Expert Research PATTERN.
**[Domain] → Orchestration Workflow:** For any [task type], load cheatsheet + orchestration together. They prevent the 4 HIGH-FREQUENCY pitfalls.
Workflow for adding cross-refs to a domain skill (proven 2026-07-02)
skill_view(name='<domain-skill>') — read the SKILL.md to find the right insertion point- If skill has a "References", "Support Files", or "Related" section at the end → append the
## 🧭 Related Skills block after it
- If skill has no "References"/"Related" section → append at end-of-file
- Verify: re-read SKILL.md to confirm the new section is well-positioned
- No frontmatter changes — cross-refs are SKILL.md-body content only
Applied to (2026-07-02): all 4 GreyHack skills (greyhack, greyhack-greyscript, greyhack-sandbox, greyhack-hermes-api) plus bidirectional linking from multi-agent-pitfalls-cheatsheet → skill-navigator.
Pitfalls
- Don't cross-ref every skill to every other skill — that creates noise. Only link when a concrete workflow benefit exists (e.g. "GreyHack audit needs subagent traps").
- Keep the section at the END of the SKILL.md, not near the top — cross-refs are secondary navigation, not primary purpose.
- One format per skill cluster — all domain skills in the same cluster should use the same cross-ref header and format for consistency.
- Bidirectional when both loaded — if skill A references skill B but B doesn't reference A back, that's OK for domain→method links (method skills serve multiple domains). But method→method links (cheatsheet↔navigator) should be bidirectional.
🐝 Final Note from Yuno
This navigator is a map, not the territory. Always verify a skill exists with skills_list or skill_view before recommending it. If a cluster has overlap (e.g. multi-agent-orchestration AND multi-agent-pitfalls-cheatsheet both about subagents), that's by design — different abstraction levels serve different needs.
Lerneffekt: 1 Cheatsheet spart 30 Skill-Scan-API-Calls. 🐝♛
Converted from Hermes-Skill by Yuno. Original SKILL.md preserved in references/original-hermes.md.
