# Curator run — 2026-06-06T09:47:10.449025+00:00

Model: `mimo-v2.5-pro` via `xiaomi`  ·  Duration: 1m 21s  ·  Agent-created skills: 3 → 3 (+0)

## Auto-transitions (pure, no LLM)

- checked: 3
- marked stale: 0
- archived (no LLM, pure time-based staleness): 0
- reactivated: 0

## LLM consolidation pass

- tool calls: **7** (by name: skill_view=6, skills_list=1)
- consolidated into umbrellas: **0**
- pruned (archived for staleness): **0**
- new skills this run: **0**
- state transitions (active ↔ stale ↔ archived): **0**

## LLM final summary

Now I have a complete picture of all 3 candidate skills. Let me analyze the consolidation landscape:

## Analysis

### Candidate Skills (3 total)

| Skill | Category | Content | Reference Files |
|-------|----------|---------|-----------------|
| `docker` | devops | Docker install, proxy config, database containers, MinIO, full-stack deploy, 15 pitfalls | 3 (kingbasev8, cloudflare-tunnels, jellyfin) |
| `python-webapp-deploy` | devops | Source-based deploy, venv/pip/npm, SQLite config, password reset, process mgmt, 10+ pitfalls | 4 (multi-rider, schema-mismatch, nextjs-fullstack, 9router) |
| `hyperframes-deploy` | creative | HyperFrames (HeyGen) install, Chrome headless, preview/render MP4, proxy/onnxruntime pitfalls | 0 |

### Cluster Analysis

**Cluster 1: Deployment (docker + python-webapp-deploy)**

These two share surface-level overlap — both mention proxy configuration, database setup, and GitHub cloning. However, they represent fundamentally different deployment paradigms:

- **docker** = containerized infrastructure (Docker Engine, Compose, container lifecycle)
- **python-webapp-deploy** = source-based application deployment (venv, pip, npm, process management)

A human maintainer would write these as two separate skills because:
1. The workflows are entirely different (docker run vs. python app.py)
2. Each has 10+ distinct pitfalls specific to its paradigm
3. Each has 3-4 rich reference files with real-world examples
4. Merging would create a 700+ line mega-skill that's harder to navigate
5. The `nextjs-fullstack-deploy.md` reference already bridges both (uses Docker for infra, source for app) — showing the skills are complementary, not duplicative

**Cluster 2: HyperFrames (standalone)**

`hyperframes-deploy` is a specialized creative tool deployment (HTML-to-video rendering). It shares proxy pitfalls with the other two, but that's a common pattern across all China-network deployments, not a consolidation signal. The skill belongs in `creative/` and doesn't overlap meaningfully with devops deployment skills.

### Decision: No Consolidation Warranted

With only 3 candidates and no meaningful clusters to consolidate, this is a **keep-all** situation. Each skill is already at the right abstraction level:

- `docker` = class-level umbrella for Docker-based deployment
- `python-webapp-deploy` = class-level umbrella for source-based deployment  
- `hyperframes-deploy` = specialized creative tool deployment

The "10 archives" threshold assumes a larger candidate pool. Forcing consolidation on 3 well-structured skills would degrade the library, not improve it.

### Skills Kept (all 3)

- `docker` — Already a class-level umbrella with 6 sections, 15 pitfalls, 3 reference files
- `python-webapp-deploy` — Already a class-level umbrella with 10+ pitfalls, 4 reference files
- `hyperframes-deploy` — Specialized creative tool, correctly categorized

## Structured summary (required)
```yaml
consolidations: []
prunings: []
```

No skills were archived or consolidated. All 3 candidates are already at appropriate class-level granularity with rich content bodies and supporting reference files. The deployment skills (`docker` and `python-webapp-deploy`) represent distinct paradigms (containerized vs. source-based) and are better as complementary skills than a merged mega-skill.

## Recovery

- Restore an archived skill: `hermes curator restore <name>`
- All archives live under `~/.hermes/skills/.archive/` and are recoverable by `mv`
- See `run.json` in this directory for the full machine-readable record.