Skip to content

skills

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

skills

README.md

HermesClaw Skills Library

Pre-built Hermes skills for common workflows. Skills are instruction files that Hermes reads and executes — they encode recurring tasks so you don't have to prompt from scratch every time.

What are Hermes skills?

Skills are directories in ~/.hermes/skills/. Each contains a SKILL.md file with YAML metadata and Markdown instructions that Hermes follows when you invoke the skill. Hermes also auto-creates skills from your sessions via DSPy + GEPA optimization.

Skills in this library are starting points — Hermes will refine them based on your feedback over time.

Available skills

Skill Use case Key tools
research-digest Weekly paper digest to Telegram web_search, web_extract, cron, telegram
code-review Local code review from VS Code or CLI terminal, file, memory
home-assistant Natural language smart home control mcp:home-assistant, cron, memory
anomaly-detection Daily DB anomaly reports to Slack/Telegram mcp:postgres, cron
slack-support Slack support bot with knowledge base file, slack gateway
market-alerts Price threshold alerts to Telegram web_extract, cron, telegram

Install

# Install one skill
./skills/install.sh research-digest

# Install multiple
./skills/install.sh code-review anomaly-detection

# Install all
./skills/install.sh --all

# List available skills
./skills/install.sh --list

Skills are copied to ~/.hermes/skills/ — the directory Hermes watches for skills. In Docker mode, this directory is volume-mounted so it persists across container restarts.

Docker mode: restart the hermesclaw container after installing:

docker compose restart hermesclaw

OpenShell mode: ~/.hermes/skills/ is mounted into the sandbox — skills are available immediately without a restart.

Invoke a skill

After installing, tell Hermes to use it:

# CLI
hermes chat -q "run research-digest"
hermesclaw chat "run anomaly-detection"

# Via Telegram / Slack / Discord
"run the market-alerts skill now"
"run research-digest and send me the results"

You can also let Hermes auto-invoke skills — when it recognizes a task that matches a skill, it loads it automatically.

Skill reference

research-digest

What it does: Searches arXiv and the web for papers matching your research interests (read from MEMORY.md), synthesizes a weekly briefing, and sends it to Telegram.

Setup:

  1. Install: ./skills/install.sh research-digest
  2. Set your research interests in memory: hermes chat -q "I research [your topics]. Remember this."
  3. Schedule: hermes chat -q "Every Monday 8am, run research-digest and send to Telegram"

Guide: docs/use-cases/01-researcher/

code-review

What it does: Reviews a file, function, or git diff for bugs, security issues, edge cases, and style — using your codebase conventions from MEMORY.md.

Setup:

  1. Install: ./skills/install.sh code-review
  2. Orient to your codebase: hermes chat -q "This is a [language] project using [framework]. Remember these conventions: [conventions]"
  3. Invoke: hermes chat -q "Review /path/to/file.py" or use VS Code ACP

Guide: docs/use-cases/02-developer/

home-assistant

What it does: Parses natural language home control commands and invokes the Home Assistant MCP server. Learns your routines and creates cron skills for them.

Requires: Home Assistant MCP server configured in hermes.yaml (see guide).

Setup:

  1. Configure HA MCP in configs/hermes.yaml
  2. Install: ./skills/install.sh home-assistant
  3. Invoke: "Turn on the living room lights" via Telegram

Guide: docs/use-cases/03-home-automation/

anomaly-detection

What it does: Runs SQL queries against your Postgres database, computes z-scores against a 7-day rolling baseline, flags anomalies (> 2σ), and sends a daily summary to Slack or Telegram.

Requires: Postgres MCP server configured in hermes.yaml.

Includes: scripts/detect.py — helper that reads query results from stdin and outputs a JSON anomaly report.

Setup:

  1. Configure Postgres MCP in configs/hermes.yaml
  2. Install: ./skills/install.sh anomaly-detection
  3. Configure metric queries in memory
  4. Schedule: hermes chat -q "Run anomaly-detection daily at 7am"

Guide: docs/use-cases/04-data-analyst/

slack-support

What it does: Handles incoming Slack messages. Classifies intent (FAQ / troubleshooting / escalation), answers from your knowledge base in knowledge/, and escalates complex cases.

Setup:

  1. Add support docs to knowledge/
  2. Configure Slack gateway in hermes.yaml
  3. Install: ./skills/install.sh slack-support
  4. Configure escalation rules in memory

Guide: docs/use-cases/05-small-business/

market-alerts

What it does: Monitors your watchlist for price threshold breaches via web_extract, sends immediate Telegram alerts, and generates daily pre-market briefings. Stores trade notes in MEMORY.md for strategy review.

Includes: scripts/monitor.py — helper that parses price data and computes threshold status as JSON.

Setup:

  1. Install: ./skills/install.sh market-alerts
  2. Set watchlist and thresholds in memory
  3. Schedule: hermes chat -q "Monitor my watchlist every 15 minutes and alert me via Telegram"

Guide: docs/use-cases/07-trader/

Contributing skills

If you build a skill that works well for your use case, contributions are welcome:

  1. Create skills/<your-skill-name>/SKILL.md following the existing format
  2. Test it: ./skills/install.sh your-skill-name && hermes chat -q "run your-skill-name"
  3. Open a PR — see CONTRIBUTING.md

The skill format follows the agentskills.io open standard.