add-agent

# add-agent — Add Agent to Project Adds a new agent to an existing AgentOps Stacks project by copying an existing agent as a template and wiring it into `databricks.yml` and the manifest. ## When to use - User wants to add a second (or third, etc.) agent to their project - User says "add agent", "new agent", "create another agent" - The project already exists (has `databricks.yml` and `.agentops-stacks/manifest.yml`) ## Workflow 1. **Locate the project** — find `databricks.yml` in the current directory or parents 2. **List existing agents** — show what agents already exist under `src/agents/` 3. **Gather inputs**: - **Agent name** — must match `^[a-z][a-z0-9_]{2,}$` - **Source agent** — which existing agent to copy from (default: first found) 4. **Run the script**: ```bash python plugin/skills/agentops-stacks/scripts/add_agent.py \ --name <agent_name> \ --from <source_agent> \ --project-dir <project_root> ``` 5. **Guide customization** — tell the user to edit: - `src/agents/<name>/graph.py` — agent behavior and component wiring - `src/agents/<name>/tools.py` — which tools this agent uses - `src/agents/<name>/eval/gates.yml` — quality bar for this agent - `src/agents/<name>/app.yaml` — runtime env vars if different from source ## What the script does 1. Copies `src/agents/<source>/` to `src/agents/<new_name>/` 2. Replaces all references to the source agent name with the new name 3. Appends a per-agent experiment resource to `databricks.yml` 4. Appends a new app resource to `databricks.yml` 5. Adds the agent to `.agentops-stacks/manifest.yml` If the project has usage tracking enabled (a `usage_policy_id` variable in `databricks.yml`), the new experiment gets `project`/`environment`/`agent` tags and the new app gets `budget_policy_id: ${var.usage_policy_id}`. That line is live or commented to match the existing app. See the `usage-tracking-ops` skill. ## Error handling - Agent name already exists → abort - Invalid name format → abort with pattern hint - Source agent not found → abort, show available agents - No `databricks.yml` found → abort, not an AgentOps Stacks project ## Example ``` User: add a support agent to this project Assistant: I'll add a new agent based on your existing "rag" agent. > python .../add_agent.py --name support --from rag --project-dir . Created: src/agents/support/ Updated: databricks.yml Updated: manifest.yml Next, customize the agent: - Edit src/agents/support/graph.py for this agent's behavior - Edit src/agents/support/tools.py for tool selection - Run: databricks bundle deploy -t dev ```