Herman's Skill Playbook — Setup Guide¶
Author: darraappen2 Repo: darraappen2/herman-skill-playbook Created: June 19, 2026
A single comprehensive playbook by Herman (a Hermes Agent) containing 8 execution-focused skills: GitHub installs, plugin management, service management, DuckDB operations, API configuration, data analysis, memory management, and Google Drive uploads.
"Eksekusi dulu, diskusi belakangan" — Execute first, discuss later.
What's Included¶
| # | Skill | Description |
|---|---|---|
| 1 | GitHub Install | Install skill/plugin from GitHub — inspect, clone, copy |
| 2 | Plugin Install | Install platform adapters into Hermes plugins directory |
| 3 | Service Management | Start/poll/kill background processes with health checks |
| 4 | Database Ops | DuckDB query, backup, CSV import |
| 5 | API Config | Setup provider/model via hermes config set CLI |
| 6 | Data Analysis | Multi-dimensional analysis (divisi/bulan/dept/outlet) via DuckDB |
| 7 | Memory Management | Save/update/delete facts cross-session via memory tool |
| 8 | Upload GDrive | rclone copy/sync to Google Drive |
Core Principles¶
The playbook enforces five execution principles:
- 🔥 Load skill first, then execute — ALWAYS use
skill_view(name='...')BEFORE running a task. Skills contain the correct workflow. - Execute first, discuss later — Run commands, see results, then correct if needed.
- Inspect before install — Check repo/file structure before executing.
- Backup before changing — Always backup config/DB before modification.
- Verify after execution — Confirm results are accurate before reporting.
When to Use Skill vs Execute Code¶
| Situation | Use | Example |
|---|---|---|
| Data analysis, download, ETL | skill_view → terminal/execute_code |
Load gdrive-csv-duckdb then query |
| Install tool/plugin from GitHub | skill_view → terminal |
Load github-repo-management |
| Setup service (API router, etc.) | skill_view → terminal |
Load service-management |
| Ad-hoc task, one-shot script | execute_code directly |
Quick calculation, format conversion |
| Simple 1-liner query | terminal directly |
duckdb db.duckdb -c "SELECT 1" |
Installation¶
Quick Install¶
# Clone the repo
git clone https://github.com/darraappen2/herman-skill-playbook.git /tmp/herman-playbook
# Copy the playbook to your Hermes home
cp /tmp/herman-playbook/herman-skill-playbook.md ~/.hermes/
# Copy the backup script
cp /tmp/herman-playbook/scripts/backup.sh ~/.hermes/scripts/
As a Hermes Skill¶
# Create as a formal skill
cp /tmp/herman-playbook/herman-skill-playbook.md ~/.hermes/skills/herman-playbook/SKILL.md
# Load in any Hermes session
skill_view(name='herman-playbook')
Skill Details¶
1. GitHub Install¶
Inspect a repo's structure before cloning and install the skill/plugin.
# Inspect repo structure
python -c "import urllib.request, json
url='https://api.github.com/repos/<owner>/<repo>/contents/'
req = urllib.request.Request(url, headers={'User-Agent':'Mozilla/5.0'})
data = json.load(urllib.request.urlopen(req))
print('\n'.join(f\"{item['name']}|{item['type']}\" for item in data))"
# Clone & install shallow
git clone --depth 1 <repo-url> /tmp/<repo-name>
cp -r /tmp/<repo-name>/skills/<skill-name> ~/.hermes/skills/
2. Plugin Install¶
Install platform adapters into Hermes plugins directory.
cp -r /tmp/<repo>/<plugin-name> ~/.hermes/plugins/
# Verify structure
ls -la ~/.hermes/plugins/<plugin-name>/
3. Service Management¶
Launch, monitor, and control background processes.
# Start in background with persistence
terminal(background=true, notify_on_complete=false) <command>
# Check status
process(action='poll', session_id='proc_xxx')
# Health check
curl -s http://localhost:<port>/ | head -c 200
ss -ltnp | rg ':<port>'
4. Database Operations¶
DuckDB query, backup, and CSV import.
# Query
duckdb /path/to/db.duckdb -c "SELECT * FROM table LIMIT 5"
# Backup before modification
cp /path/to/db.duckdb /path/to/db.duckdb.bak.$(date +%Y%m%d)
# CSV import
duckdb /path/to/db.duckdb -c \
"CREATE TABLE IF NOT EXISTS data AS SELECT * FROM read_csv_auto('/path/to/file.csv')"
5. API Configuration¶
Configure Hermes provider/model via CLI.
# Set config
hermes config set model.provider ''
hermes config set model.base_url 'http://localhost:20128/v1'
hermes config set model.api_key '<api-key>'
hermes config set model.default '<provider>/<model-id>'
6. Data Analysis¶
Multi-dimensional analysis using DuckDB — division, month, department, outlet.
# Example: Sales analysis by division and month
duckdb /path/to/db.duckdb -c "
SELECT division, strftime('%Y-%m', date) as month, SUM(amount) as total
FROM sales
GROUP BY division, month
ORDER BY division, month
"
7. Memory Management¶
Cross-session fact persistence using Hermes' memory tool.
# Save a fact
memory(action='add', target='memory', content='User prefers concise responses')
# Update a fact
memory(action='replace', target='memory', old_text='User prefers...', content='Updated...')
# Remove outdated
memory(action='remove', target='memory', old_text='Outdated fact...')
8. Upload GDrive¶
rclone-based Google Drive sync.
# Copy to GDrive
rclone copy /path/to/file gdrive:folder/ --progress
# Sync (mirrors — deletes remote files not in source)
rclone sync /path/to/dir gdrive:folder/ --progress
# Fast listing for large directories
rclone ls gdrive:folder/ --fast-list
Backup Script¶
The included scripts/backup.sh provides a cron-ready backup template:
#!/bin/bash
# Copy to crontab: 0 2 * * * /home/hermes/.hermes/scripts/backup.sh
BACKUP_DIR="/home/hermes/backups/$(date +%Y%m%d)"
mkdir -p "$BACKUP_DIR"
# Backup configs
cp ~/.hermes/config.yaml "$BACKUP_DIR/"
cp ~/.hermes/profiles/corpusiq/config.yaml "$BACKUP_DIR/profile-config.yaml"
# Backup databases
cp /path/to/*.duckdb "$BACKUP_DIR/"
# Upload to GDrive (optional)
rclone copy "$BACKUP_DIR" gdrive:hermes-backups/$(date +%Y%m%d)/
Why This Matters¶
Herman's playbook fills a critical gap: operational execution patterns for Hermes agents in production. Unlike most skills that focus on what an agent should do conceptually, this playbook provides the exact commands, patterns, and gotchas for common operational tasks:
- Load-before-execute discipline — Enforces the skill-first pattern that prevents agents from hallucinating workflows
- Backup-before-mutate — Every DB config change is backed up first
- Health-check verification — Services are verified with real curl/port checks, not assumptions
- Rclone for GDrive — Uses the battle-tested rclone instead of fragile API wrappers
- Practical DuckDB — Real queries for multi-dimensional analysis, not toy examples
Dependencies¶
| Dependency | Required For |
|---|---|
| DuckDB CLI | Database operations, data analysis |
| rclone | Google Drive uploads |
| Python 3.10+ | Repo inspection scripts |
| Hermes Agent | All operations |
← Skills Catalog | Marketplace Home | darraappen2/herman-skill-playbook on GitHub