Watchdog / heartbeat: anti-stall for long autonomous runs (design spec)#102
Merged
Conversation
Design spec (RFC) for an orchestrator-owned watchdog that closes the idle-stall hole: an event-driven lead receives no wake signal when every teammate goes idle or a background job dies silently, so a run can flat-line unnoticed. Spec covers a scheduled heartbeat tick, evidence-based liveness (pgrep + artifact deltas, not absence-of-bad-news), a long-job failure-aware monitor, remediation escalation (nudge -> respawn -> reroute), config, and integration points. Implementation follows in a separate PR. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Deploying zero-operators with
|
| Latest commit: |
81cb101
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://a58bf737.zero-operators.pages.dev |
| Branch Preview URL: | https://feat-agent-watchdog-heartbea.zero-operators.pages.dev |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a design spec (RFC) for an orchestrator-owned watchdog / heartbeat that closes an idle-stall hole in long autonomous runs.
specs/watchdog.md.Why
The lead-orchestrator is event-driven — it acts on messages from the human or teammates. That has a silent-death hole: when every teammate goes idle (or a spawn dies silently), no message is emitted, so the orchestrator gets no wake signal and is never re-invoked. A long background job (training run, sweep, external CI) emits no agent message while it runs either. Net: "wait for the next message" degrades into "do nothing, indefinitely" — the run looks alive but has flat-lined. This was observed in the wild: a full team stalled silently for ~38h before the human noticed.
This is distinct from context saturation (already handled by checkpoint→respawn). This is idle-stall.
Design (summary)
*/17 * * * *), re-armed each session, independent of teammate messages.Scope
Spec-only. Code + tests + orchestrator/comms wiring land in a follow-up implementation PR against this spec.
Test
N/A (spec). Acceptance criteria for the implementation are enumerated in §5.
🤖 Generated with Claude Code