Hermes Module

Generates the quick_commands block in ~/.hermes/config.yaml and a full r-research SKILL.md from .github/prompts/*.prompt.md, the source of truth for all slash commands.

Overview

Hermes has no native prompt-file format at all — it's driven entirely by ~/.hermes/config.yaml (zero-token quick_commands) and skill files. sync.py reads every prompt file via modules/common/prompt_commands.py, classifies each one (exec / exec_long / arg / ai_guided) based on its own Hermes-specific rules (e.g. which commands are too long-running for the 30s quick_commands timeout), and writes both output files.

claude is skipped — not useful as a Hermes /r-* command.

Usage

uv run --no-sync python -m modules.hermes.sync
uv run --no-sync invoke hermes.sync

Never hand-edit ~/.hermes/config.yaml's quick_commands key or the generated SKILL.md — re-run this sync after adding or modifying any .github/prompts/*.prompt.md file, then start a new Hermes session (or /reset) to pick up the changes.

Files

sync.py — reads .github/prompts/ via modules/common/prompt_commands.py, classifies commands, and writes ~/.hermes/config.yaml + ~/.hermes/skills/r-research/SKILL.md
README.md — this file

Architecture

uv run --no-sync invoke hermes.sync
  ↓
modules/hermes/sync.py
  ↓
modules/common/prompt_commands.py (shared .prompt.md parser)
  ↓
~/.hermes/config.yaml (quick_commands) + ~/.hermes/skills/r-research/SKILL.md

FilesExpand file tree

README.md

Latest commit

History