Shareable configuration for OpenClaw AI assistant — templates, memory architecture, and integration skills.
- Skills: Standalone UV scripts (Python 3.11+, inline dependencies)
- Tests: pytest via
uv run --with pytest pytest tests/ -v - No project-level dependencies — each skill is self-contained
skills/— Integration CLIs (Parallel, Limitless, Fireflies, Quo, OpenClaw)templates/— AGENTS.md, SOUL.md, USER.md templates for OpenClaw instancesmemory/— Example memory architecture structuretests/— Skill tests (integration tests auto-skip without API keys)
DO:
- Skills are standalone UV scripts with
# /// scriptinline dependencies - Each skill has
SKILL.md(metadata) + executable script (same name as directory) - Bump
VERSIONfile and skill'sSKILL.mdversion on changes - Tests skip gracefully when API keys unavailable
DON'T:
- Add pyproject.toml — this is not a Python package
- Share code between skills — each is self-contained
- Commit API keys or
.envfiles
This repo is the upstream source for live OpenClaw instances. The openclaw skill
handles installation and updates:
templates/→ copied to instance workspace on install/updateskills/→ copied to instance workspace (safe to overwrite)workflows/→ onlyAGENT.mdupdates;rules.md,agent_notes.md,logs/are user-owned and never overwrittendevops/→ health check and machine setup specs, deployed via cron on fleet machines
This matters: skills and templates can be freely edited here, but workflow files besides
AGENT.md belong to the running instance and must not be clobbered on update.
This project has been renamed twice: Clawdbot → Moltbot → OpenClaw (current,
after Anthropic trademark concerns). Legacy paths and config references using old names
still exist — both work, but prefer openclaw going forward.
Commit style: Conventional or emoji prefix, Co-Authored-By for AI commits.
Example: Rewrite Parallel.ai skill from Bash to Python