cwc-workshops

Learn to implement dispatch handlers in Anthropic CWC workshops. Fan out complex tasks to specialized agents using a central dispatch router from tool names to Python functions.

Learn to implement memory mount for shared knowledge in analyst sessions by supplying a memory_store resource with read_write access when creating your session.

Implement scorecard validation for agent output quality using the validateScorecard function. Enforce schema compliance and get detailed feedback on validation failures.

Stream session events from Claude agents in real-time using client.beta.sessions.events.stream(). Handle agent messages, tool calls, and status updates as they happen.

Learn to create sub agents that delegate tasks to specialist agents using Anthropic API and orchestrator tools. Optimize task delegation for parallelizable and context-heavy workloads.

Learn how to integrate HashiCorp Vault for secure credential handling in Claude agents. Eliminate hard-coded secrets and leverage built-in secret management with CWC Workshops.

Evaluate Claude Managed Agents with decision probes in the CWC agent-battle harness. This technical guide details using synthetic Minecraft states for rapid, rubric-based scoring.

Learn how to use memory stores for cross-session persistence in Claude agents. Mount a CMA Memory Store as a virtual file system for shared, persistent data across chat sessions.

Learn the workflow for versioned agent updates and skill publishing in the anthropics/cwc-workshops repository. Discover idempotent deployment for Claude Managed Agents.

Learn how to implement outcome evaluation with file-based rubrics for grading in CWC workshops. Use YAML rubrics and an LLM judge to assess agent responses effectively.

Learn how MCP servers connect to Claude Managed Agents using the vault credential pattern. Discover how static bearer vault credentials bind relay secrets to server URLs.

Understand the Scorecard outcome grading system with hybrid evaluation. Learn how programmatic and LLM-as-judge methods assess agent task completion for standardized results.