skill(writing-agent-relay-workflows): lessons from multi-CLI game test#56
skill(writing-agent-relay-workflows): lessons from multi-CLI game test#56khaliqgant wants to merge 3 commits into
Conversation
… test Four additions surfaced by running a real multi-turn, multi-CLI workflow end-to-end (claude/codex/opencode peer guessing game): - New coordination shape "Per-turn interactive spawn" — interactive PTY per turn rather than one long-running session. Avoids the idle-detection-vs-wait-for-event race that bites long-running interactive agents in turn-taking flows. - New caveat at top of Chat-Native Coordination Recipes: the runtime's ~30s idle threshold beats "wait for X" / "exit on GAME_OVER" task prompts. Every chat-native recipe encodes a wait, so this affects all of them. Includes the log symptom to watch for. - New Common Mistakes row: making an LLM agent the judge/oracle for fixed-rule comparisons. Observed in the wild: claude-as-oracle drove timing with shell `sleep 30` calls because PTY agents have no real wait API. If the comparison is computable, grade deterministically. - Counterweight to "Workflows Repair Before They Fail": skip the repair step when the gate sits downstream of deterministic work and has no agent-fixable failure mode. Otherwise the workflow burns a spawn cycle on a guaranteed no-op. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…e hatch Augment the new Chat-Native Coordination Recipes idle-detection caveat with two explicit handling options: (1) per-agent idleThresholdSecs override (already supported by the builder; default 30s; 0 disables), and (2) restructure to Per-turn interactive spawn. (1) is the lower-friction option for the common case "this one reviewer/grader waits long stretches on channel events" — no DAG reshape needed. (2) is the right answer when the whole flow is turn-taking. Pairs with AgentWorkforce/relay JSDoc expansion that makes idleThresholdSecs discoverable in the SDK builder docs. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
Warning Rate limit exceeded
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
…ssons Bumps the writing-agent-relay-workflows package version to match the content additions in this branch (per-turn coordination shape, idle- detection caveat, LLM-as-oracle anti-pattern, repair-step counterweight, idleThresholdSecs escape-hatch documentation). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
1 participant
Summary
Four additions to the
writing-agent-relay-workflowsskill, surfaced by running a real multi-turn workflow end-to-end acrossclaude,codex, andopencodeas peer players.New coordination shape: Per-turn interactive spawn. A third row in the Conversation/Pipeline table. Interactive PTY agents (no preset), but each step is bounded to one turn that exits when its file/message is produced. Use when turns are short and well-defined; avoids the idle-detection-vs-wait-for-event race that bites long-running interactive agents.
Idle-detection caveat at top of Chat-Native Coordination Recipes. The runtime's ~30s idle threshold beats every recipe below — Q/A, Broadcast/Ack, Peer Review, Hand-Off all encode a wait. Two handling options:
idleThresholdSecson the agent (default 30s;0disables). Pairs with the SDK JSDoc expansion in AgentWorkforce/relay#TBD.New Common Mistakes row: LLM-as-oracle anti-pattern. Observed in the wild: an LLM oracle drove timing with shell
sleep 30calls because PTY agents have no real wait API. If the comparison is computable (guess vs secret, test pass/fail, regex match), grade it deterministically. LLMs should produce, not adjudicate fixed rules.Counterweight to "Workflows Repair Before They Fail." Skip the repair step when the gate sits downstream of deterministic work and has no agent-fixable failure mode. Otherwise the workflow burns a spawn cycle (and often a follow-on idle-timeout wait) on a guaranteed no-op.
Test plan
AgentWorkforce/workflows—tests/guessing-game/workflow.ts— which exhibited all four issues before the fixes.skills/writing-agent-relay-workflows/SKILL.mdmodified (+21 lines).🤖 Generated with Claude Code