# Curator run — 2026-06-11T15:41:14.703641+00:00 Model: `gpt-5.5` via `openai-codex` · Duration: 11m 28s · Agent-created skills: 61 → 13 (-48) ## Auto-transitions (pure, no LLM) - checked: 61 - marked stale: 0 - archived (no LLM, pure time-based staleness): 0 - reactivated: 0 ## LLM consolidation pass - tool calls: **93** (by name: skill_manage=71, skill_view=4, skills_list=1, terminal=17) - consolidated into umbrellas: **2** - pruned (archived for staleness): **46** - new skills this run: **0** - state transitions (active ↔ stale ↔ archived): **0** ### Consolidated into umbrella skills (2) _These skills were **absorbed into another skill** during this run — their content still lives, just under a different name. The original directory was moved to `~/.hermes/skills/.archive/` for safety and can be restored via `hermes curator restore ` if the consolidation was wrong._ - `kanban-orchestrator` → merged into `hermes-agent` — Kanban orchestration is a Hermes feature role and belongs in the Hermes operations umbrella. - `kanban-worker` → merged into `hermes-agent` — Kanban worker behavior is a Hermes feature role and belongs beside orchestrator guidance. ### Pruned — archived for staleness (46) _These skills were archived without being merged into an umbrella (e.g. stale, unused, or judged irrelevant). Directories live under `~/.hermes/skills/.archive/`. Restore any via `hermes curator restore `._ - `airtable` - `architecture-diagram` - `arxiv` - `ascii-art` - `ascii-video` - `astro-marketing-websites` - `audiocraft-audio-generation` - `baoyu-infographic` - `blogwatcher` - `claude-design` - `design-md` - `edge-tts` - `evaluating-llms-harness` - `excalidraw` - `gif-search` - `google-workspace` - `himalaya` - `huggingface-hub` - `llama-cpp` - `llm-wiki` - `local-business-website-redesign` - `manim-video` - `maps` - `nano-pdf` - `notion` - `obsidian` - `ocr-and-documents` - `p5js` - `polymarket` - `popular-web-designs` - `powerpoint` - `pretext` - `research-paper-writing` - `segment-anything-model` - `serving-llms-vllm` - `sketch` - `social-media-shorts-automation` - `teams-meeting-pipeline` - `touchdesigner-mcp` - `vault-architecture` - `web-scraping` - `weights-and-biases` - `xurl` - `yazio-integration` - `youtube-content` - `yuanbao` ## LLM final summary Completed an umbrella-building consolidation pass. Summary of actions: - Created new class-level umbrella skills: - `visual-artifact-design` - `website-production` - `research-workflows` - `document-productivity-workflows` - `knowledge-vault-workflows` - `llm-model-operations` - `workspace-api-integrations` - `media-content-utilities` - `social-platform-operations` - Patched existing umbrellas: - `hermes-agent` now includes the absorbed Kanban orchestrator/worker playbooks. - `health-management` now includes the absorbed YAZIO nutrition integration. - Preserved package integrity: - Source packages with support files were copied into the target umbrella under `references/absorbed-packages//`. - Copied package `SKILL.md` files were renamed to `SOURCE_SKILL.md` so they do not reappear as active skill roots. - Added `references/absorbed-packages-index.md` files to the new umbrellas for discoverability. - Archived consolidated source skills under `.archive/`; where `skill_manage delete` removed a package instead of leaving a visible archive, I reconstructed the archive from the preserved absorbed-package copy. - Verified active skill discovery after consolidation: active skill list no longer exposes the absorbed siblings or nested copied packages as active skills. Clusters processed: - Creative artifact/design/tooling cluster → `visual-artifact-design` - Website/marketing/local-business cluster → `website-production` - Research/discovery/scraping/paper-writing cluster → `research-workflows` - Document/PDF/OCR/slides/Teams pipeline cluster → `document-productivity-workflows` - Obsidian/vault architecture cluster → `knowledge-vault-workflows` - MLOps/model serving/evaluation/hub/model utility cluster → `llm-model-operations` - Workspace/SaaS/API integration cluster → `workspace-api-integrations` - YouTube/GIF/TTS media utility cluster → `media-content-utilities` - Social publishing/platform integration cluster → `social-platform-operations` - Hermes Kanban role split → `hermes-agent` - YAZIO nutrition integration → `health-management` ## Structured summary (required) ```yaml consolidations: - from: architecture-diagram into: visual-artifact-design reason: Diagram generation is a visual artifact subtype better discovered under one visual-design umbrella. - from: claude-design into: visual-artifact-design reason: One-off HTML artifact design belongs with the broader visual artifact workflow. - from: popular-web-designs into: visual-artifact-design reason: Design-system reference patterns support visual artifact creation rather than needing a standalone trigger. - from: sketch into: visual-artifact-design reason: Throwaway HTML mockups are a subsection of visual artifact exploration. - from: p5js into: visual-artifact-design reason: Generative p5.js sketches are one medium within code-based visual artifacts. - from: excalidraw into: visual-artifact-design reason: Excalidraw diagrams are a hand-drawn diagram subtype under visual artifacts. - from: design-md into: visual-artifact-design reason: Design token specs support visual/system design workflows under the same umbrella. - from: baoyu-infographic into: visual-artifact-design reason: Infographic production is a visual artifact subtype with preserved layout/style details. - from: ascii-art into: visual-artifact-design reason: ASCII art is a terminal/text visual artifact subsection. - from: ascii-video into: visual-artifact-design reason: ASCII video is a media-specific visual artifact subsection. - from: manim-video into: visual-artifact-design reason: Manim educational animation is an animation subtype within visual artifacts. - from: pretext into: visual-artifact-design reason: Pretext demos are typographic/browser visual artifacts and fit the shared class. - from: touchdesigner-mcp into: visual-artifact-design reason: TouchDesigner realtime visuals are a specialized visual artifact toolchain. - from: astro-marketing-websites into: website-production reason: Astro marketing sites and local-business sites share the maintainable website production workflow. - from: local-business-website-redesign into: website-production reason: Local business redesign is a website-production subtype with SEO/content specifics. - from: arxiv into: research-workflows reason: arXiv search is a literature-discovery subsection of the research workflow. - from: blogwatcher into: research-workflows reason: Blog/feed monitoring is a recurring research-intake subsection. - from: llm-wiki into: research-workflows reason: Building/querying a markdown knowledge base is downstream research organization. - from: research-paper-writing into: research-workflows reason: Paper writing belongs with discovery and synthesis as a research pipeline stage. - from: web-scraping into: research-workflows reason: Web scraping is a research/data-extraction subsection with preserved extraction tactics. - from: polymarket into: research-workflows reason: Prediction-market querying is a market/event research data-source subsection. - from: nano-pdf into: document-productivity-workflows reason: PDF editing is a document workflow subtype. - from: ocr-and-documents into: document-productivity-workflows reason: OCR and document extraction are core document productivity operations. - from: powerpoint into: document-productivity-workflows reason: Slide deck creation/editing is an office-document workflow subtype. - from: teams-meeting-pipeline into: document-productivity-workflows reason: Meeting summaries and Teams pipeline operations produce document-style office artifacts. - from: obsidian into: knowledge-vault-workflows reason: Operational Obsidian note work belongs with local knowledge-vault workflows. - from: vault-architecture into: knowledge-vault-workflows reason: Vault structure design is a subsection of knowledge-vault management. - from: huggingface-hub into: llm-model-operations reason: Hugging Face model/data handling is a model-operations stage. - from: llama-cpp into: llm-model-operations reason: Local GGUF inference is one inference subtype in model operations. - from: serving-llms-vllm into: llm-model-operations reason: vLLM serving is the production-serving subsection of model operations. - from: evaluating-llms-harness into: llm-model-operations reason: Benchmarking and evaluation are core model-operations activities. - from: weights-and-biases into: llm-model-operations reason: Experiment tracking supports the shared model-operations lifecycle. - from: segment-anything-model into: llm-model-operations reason: Specialized model utility workflows fit the broader model-operations umbrella. - from: audiocraft-audio-generation into: llm-model-operations reason: AudioCraft model execution is a specialized model-operations subtype. - from: kanban-orchestrator into: hermes-agent reason: Kanban orchestration is a Hermes feature role and belongs in the Hermes operations umbrella. - from: kanban-worker into: hermes-agent reason: Kanban worker behavior is a Hermes feature role and belongs beside orchestrator guidance. - from: google-workspace into: workspace-api-integrations reason: Google Workspace automation is one SaaS/API workspace integration. - from: airtable into: workspace-api-integrations reason: Airtable CRUD/upsert workflows are a workspace API subsection. - from: notion into: workspace-api-integrations reason: Notion page/database automation is a workspace API subsection. - from: himalaya into: workspace-api-integrations reason: IMAP/SMTP email operations are workspace communication API workflows. - from: maps into: workspace-api-integrations reason: Geocoding/routing/timezone utilities are supporting workspace API lookups. - from: youtube-content into: media-content-utilities reason: YouTube transcript/content extraction is a media utility workflow. - from: gif-search into: media-content-utilities reason: GIF search/download is a small media retrieval utility. - from: edge-tts into: media-content-utilities reason: Text-to-speech generation is a media/audio utility workflow. - from: social-media-shorts-automation into: social-platform-operations reason: Short-form publishing automation is a major subsection of social platform operations. - from: xurl into: social-platform-operations reason: X/Twitter API actions are platform operations under social workflows. - from: yuanbao into: social-platform-operations reason: Yuanbao group interactions are social/messaging platform operations. - from: yazio-integration into: health-management reason: YAZIO nutrition retrieval is a health-data subsection for nutrition correlation. prunings: [] ``` ## Recovery - Restore an archived skill: `hermes curator restore ` - All archives live under `~/.hermes/skills/.archive/` and are recoverable by `mv` - See `run.json` in this directory for the full machine-readable record.