II.
Page JSON
Structured · livepage:docs-harness-features-backlog-gaps-subagent-observability-gap-subobs-004
GAP-SUBOBS-004: Subagent Health and Timeout Monitoring json
Inspect the normalized record payload exactly as the atlas UI reads it.
{
"id": "page:docs-harness-features-backlog-gaps-subagent-observability-gap-subobs-004",
"_kind": "Page",
"_file": "wiki/docs/harness-features-backlog/gaps/subagent-observability/gap-subobs-004.md",
"_cluster": "wiki",
"attributes": {
"nodeKind": "Page",
"sourcePath": "docs/harness-features-backlog/gaps/subagent-observability/GAP-SUBOBS-004.md",
"sourceKind": "repo-docs",
"title": "GAP-SUBOBS-004: Subagent Health and Timeout Monitoring",
"displayName": "GAP-SUBOBS-004: Subagent Health and Timeout Monitoring",
"slug": "docs/harness-features-backlog/gaps/subagent-observability/gap-subobs-004",
"articlePath": "wiki/docs/harness-features-backlog/gaps/subagent-observability/GAP-SUBOBS-004.md",
"article": "\n# GAP-SUBOBS-004: Subagent Health and Timeout Monitoring\n\n| Field | Value |\n|-------|-------|\n| Category | subagent-observability |\n| Priority | Medium |\n| Effort | M |\n| Status | Missing |\n\n## Description\nMonitor health of running subagents: detect hangs, timeouts, excessive token usage. Alert and auto-recover.\n\n## Current State\nBasic timeout on invokeHarness(). No health monitoring or auto-recovery.\n\n## Target State\nContinuous health monitoring for running subagents. Hang detection (no output for configurable period). Token budget enforcement per subagent. Auto-recovery: retry, fallback to different harness, or abort with partial results.\n\n## Dependencies\n- [GAP-SUBOBS-001](../subagent-observability/GAP-SUBOBS-001.md) -- streaming output for health signals\n- [GAP-SUBOBS-003](../subagent-observability/GAP-SUBOBS-003.md) -- token tracking for budget enforcement\n\n## Key Files\n| Component | Path |\n|-----------|------|\n| Harness invoker | `packages/sdk/src/harness/invoker.ts` |\n| Config module | `packages/sdk/src/config/` |\n\n## Recommendation\nPhase 3 implementation. Add health monitor to invokeHarness(). Detect hangs via output silence. Enforce token budgets. Implement auto-recovery strategies.\n",
"documents": []
},
"outgoingEdges": [],
"incomingEdges": [
{
"from": "page:docs-harness-features-backlog",
"to": "page:docs-harness-features-backlog-gaps-subagent-observability-gap-subobs-004",
"kind": "contains_page"
}
]
}