SKILL.md


name: lucy-agent description: > Lucy-ai's own agent ecosystem installer, updater, and verifier. Use this skill when:

  • Installing lucy-agent on a new OpenClaw instance
  • Updating lucy-agent in-place
  • Verifying the lucy-agent installation
  • Understanding the lucy-agent structure metadata: author: lucy-ai version: "1.2"

What is lucy-agent?

lucy-agent is the installable ecosystem for Lucy-ai — a fully configured AI colleague with personality, skills, conventions, and gateway config fragment.

Repository: https://github.com/camiloandresgtruniandes/lucy-agent

Installation

One-command install (fresh machine)

curl -fsSL https://raw.githubusercontent.com/camiloandresgtruniandes/lucy-agent/main/install.sh | bash

Flags

  • --skip-clawhub — Skip ClawHub skill installation
  • --skip-workspace — Skip workspace file seeding
  • --force — Overwrite conflicting files without prompting
  • --dry-run — Show what would be done without making changes

Post-install (manual)

  1. Link the config fragment in openclaw.json:
    { $include: "./lucy-agent/config/agent-fragment.json5" }
    
  2. Configure channel tokens and API keys in openclaw.json
  3. Restart OpenClaw: openclaw gateway restart

Updating

cd ~/.openclaw/lucy-agent && ./update.sh
# Or pin to a specific version:
./update.sh --tag v1.0.0

Verification

cd ~/.openclaw/lucy-agent && ./verify.sh

What's included

  • Bundled skills: sdd, github-pr, pr-review, csharp-dotnet, tailwind-4, typescript, skill-creator
  • ClawHub skills: weather, browser-automation, acp-router (auto-installed)
  • Workspace seed: SOUL.md, IDENTITY.md, AGENTS.md, USER.md, TOOLS.md, HEARTBEAT.md
  • Config fragment: agent defaults (skills, model, thinking level)

Conflict handling

lucy-agent is idempotent and careful:

  • If a workspace file was modified by the user, it prompts before overwriting
  • If a skill was modified by the user, it prompts before overwriting
  • Re-running install.sh on an existing installation does a git pull (not a fresh clone)

Security

  • No secrets are ever committed (API keys, tokens, credentials)
  • The install script is auditable — curl it and read before running
  • Config fragment contains only public configuration (skill names, model names, flags)

Troubleshooting

"OpenClaw CLI not found"

Install OpenClaw first: curl -fsSL https://openclaw.ai/install.sh | bash

verify.sh fails

Run with verbose output: bash -x verify.sh Check that all skills in skills/ have a SKILL.md file

git pull fails on existing installation

cd ~/.openclaw/lucy-agent
git status  # check for uncommitted changes
git stash   # or manually handle conflicts
git pull