setup-agents

name	setup-agents
description	Interactive setup guide to install the Gemini Agency Agents locally or globally. Trigger this when a user asks to "setup agents" or "install agents" from the repo.

Setup Agents Skill

This skill provides an interactive workflow to help a user install the Gemini Agency Agents from this repository into their Gemini CLI configuration.

Workflow Execution Steps

When triggered, you must execute the following workflow sequentially:

1. Ask for Installation Scope

Use the ask_user tool to present the user with a choice of where to install the agents:

• Header: Install Location
• Question: Where would you like to install the Agency Agents?
• Options:
  1. Global (~/.gemini/agents) (Recommended: available in all workspaces)
  2. Local (.gemini/agents) (Available only in this workspace)
  3. Other (Custom path)

2. Determine Target Directory

Based on the user's choice:

• If Global: Target directory is ~/.gemini/agents/
• If Local: Target directory is .gemini/agents/
• If Other: Use the text they provided.

Expand ~ if necessary using a shell command or by prepending the home directory path. Ensure the target directory exists by running mkdir -p <target_dir>.

3. Initialize Submodules & Build (If necessary)

Just to be safe, ensure the submodule is initialized and the python script is executed so the source files are ready. Run this via run_shell_command:

git submodule update --init --recursive
# We assume the files are already built and staged in the repo's folders, 
# but we can just copy them directly.

4. Install the Agents

Execute the bundled python installation script via run_shell_command to find all valid agent markdown files in the repository (excluding .git and agency-agents-submodule) and copy them into the chosen target directory.

Note: You must pass the target directory as an argument to the script. The script is located in your skill's scripts folder. Use the absolute path provided in your <available_resources>.

python3 <path-to-skill-creator>/scripts/install_agents.py "<TARGET_DIRECTORY_FROM_STEP_2>"

5. Instruct the User to Refresh

Once the files are copied successfully, output a message to the user instructing them to reload their agents. Output exactly this:

"I've activated the agents! Please type /skills reload and /agents reload (or simply restart your CLI session) so I can load them into memory. Once you're done, let me know and we can verify it works!"

6. Verification (Wait for User)

Wait for the user to respond that they have reloaded. Once they do, automatically trigger the agents-orchestrator to run a network check:

"I will now ask the agents-orchestrator to verify that the departments are online."

Call the agents-orchestrator tool with: "Please run a quick availability check. Call EXACTLY ONE agent from each of the following 7 departments: 1. Design & UX, 2. Engineering, 3. Marketing, 4. Product & Project Management, 5. Support & Operations, 6. Testing & Quality, and 7. Spatial Computing. Ask them: 'Are you available and online?'. Strictly instruct them NOT to execute any file operations, shell commands, or real tasks. Simply collect their natural, persona-driven responses to your question and report them back to me."

Present the orchestrator's report to the user to conclude the setup.