Repository navigation
Multi-Agent Response Standard: Structured Output for Parallel Agents #232
Description
Description
I Have ADHD is brilliant. The concept of structuring agent responses for scannability rather than completeness is exactly what multi-agent systems need. When you're running 4 agents in parallel and each one outputs walls of text, the signal-to-noise ratio drops fast.
What I Love
- "Lead with the action" — agents should say what they did, not narrate their thought process
- Shorter lists — 3 items max before summarizing
- Reduced repetition — critical when the same context is shared across agents
Suggestion: Multi-Agent Response Standard
When running multiple agents (Claude Code, etc.), each agent's output should be:
- Identified — Which agent produced this output?
- Timestamped — When? (critical for debugging async agent interactions)
- Structured — Action first, then brief context, then next steps
- Cross-referenced — If Agent A's output relates to Agent B's task, link them
Example Format:
[Agent: fix-auth] [2026-09-20 14:30]
✅ Completed: Added JWT token validation
📝 Summary: 3 files changed, 45 lines added
🔗 Related: Agent feat-api (PR #42) — depends on this auth fix
⏭️ Next: Run integration tests (agent test-suite)
Question
Is there a way to enforce this response contract across multiple agents simultaneously? Maybe a shared .adhdrc that all agents in a project inherit?
Great work — this should be standard for all AI coding agents.
Activity
rules: add multi-agent output section (issue ayghri#232)