Scenario: Code Generation with Claude Code You are using Claude Code to accelerate software development. Your team uses it for code generation, refactoring, debugging, and documentation. You need to integrate it into your development workflow with custom slash commands, CLAUDE.md configurations, and understand when to use plan mode vs direct execution. Your team has integrated Claude Code into a multi-agent CI/CD pipeline. A coordinator agent spawns two subagents: one that runs static analysis tools via Bash and one that executes the test suite. During a recent incident, the test execution subagent encountered a flaky network dependency and timed out midway through the test run. The coordinator received no structured information about what happened and defaulted to marking the entire build as failed, blocking the merge queue for 45 minutes until a developer manually investigated. A post-mortem reveals the test subagent simply propagated a raw exception to the coordinator without any context about which tests had already passed, what the failure type was, or whether a retry might succeed. What design change to the test subagent's error reporting would most directly prevent this class of coordinator paralysis?
Context Sources: A Six-Level Taxonomy
Loading lesson content...
Scenario: Multi-Agent Research System You are building a multi-agent research system using the Claude Agent SDK. A coordinator agent delegates to specialized subagents: one searches the web, one analyzes documents, one synthesizes findings, and one generates reports. The system researches topics and produces comprehensive, cited reports. During load testing, reports on broad topics become inconsistent once the research phase produces more than 30 source analyses. The document analysis subagent completes successfully, but its outputs are long narrative memos with internal reasoning, copied verbatim into the synthesis prompt. The synthesis agent omits high value evidence from earlier analyses and sometimes cites sources that were discussed but not central. What redesign best addresses the reliability problem?