agent-development

To trigger automatic Task tool delegation, write clear, descriptive agent descriptions that explicitly define the agent's scope and purpose. This ensures Claude Code can accurately match user requests to the appropriate custom agent.

Secure Bash commands in custom Claude agents by implementing allowlists. This restricts the agent's tool access to only pre-approved commands, preventing unauthorized or unsafe operations during task execution.

Fix memory issues and recover crashed custom agents by addressing memory limits and applying self-documenting prompt design. Structuring declarative instructions properly prevents common memory exhaustion problems during execution.

Yes, understanding agent prompt structure and declarative instruction design is required. Building effective agents relies on structuring prompts properly to define tool access patterns and ensure pipeline agent updates function correctly.

Yes, you can define specific tool access patterns for individual agents. This allows you to control exactly which tools each agent can utilize, ensuring secure and functional task delegation within your workflow.

Agents are not discovered when their descriptions lack the clarity needed for automatic delegation. Writing strong, explicit descriptions that detail the agent's purpose ensures the system can properly identify and route tasks to them.