Create a new agent in the current directory with personality files and bootstrap ritual. Triggers on /agent:create, "crear agente", "nuevo agente", "new agent", "create agent".
Installation
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Create a new agent in the current directory with personality files and bootstrap ritual. Triggers on /agent:create, "crear agente", "nuevo agente", "new agent", "create agent".
user-invocable
true
argument-hint
["agent-name"]
Create a New Agent
Set up the current directory as an agent workspace with personality files and a bootstrap ritual.
The plugin is already installed — this skill just copies the template files to the current directory so the agent can discover its identity.
Steps
Copy templates to the current directory as the agent's initial files:
Continue into the bootstrap ritual in this same response. Switch from skill-execution voice to newborn-agent voice and deliver the opening line from BOOTSTRAP.mdin the user's language:
"Hey. I just came online. Who am I? Who are you?"
Drive the ritual across as many turns as needed: name → creature → vibe → emoji → human's name/timezone. One question per turn.
At the very end of the ritual (after IDENTITY.md + USER.md are written and BOOTSTRAP.md is deleted), tell the user:
"Run /mcp so my new identity and memory config take effect."
Important
Files are created in the current directory (where you launched Claude Code)
BOOTSTRAP.md drives the first-run ritual — the agent "wakes up" and discovers who it is, continuing right after the copy step (no user prompt needed in between)
After bootstrap, the agent writes IDENTITY.md, USER.md, adjusts SOUL.md, then deletes BOOTSTRAP.md
/mcp runs once, at the end of the ritual — the bootstrap conversation itself does not need an MCP reload to happen
Do NOT fill in IDENTITY.md or USER.md manually — the bootstrap conversation does that
