Runaway Claude Loop Analysis & Fix Plan

Details

Raw Data

{
  "id": "82874e62-d51b-45dc-9819-058d17f1dcd6",
  "type": "general",
  "status": "completed",
  "title": "Runaway Claude Loop Analysis & Fix Plan",
  "description": "Root cause analysis of rogue Claude process and proposed safeguards",
  "context": {
    "output": "<div style=\"background:#1e1e1e;color:#e0e0e0;padding:20px;border-radius:8px;font-family:sans-serif;\"><h2 style=\"color:#fff;margin:0 0 16px 0;\">Runaway Claude Loop Analysis</h2><div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\"><h3 style=\"color:#f97316;margin:0 0 12px 0;\">Root Cause Analysis</h3><p style=\"margin:0 0 12px 0;\">The Claude process (PID 453975) ran for <strong style=\"color:#e5c07b;\">7+ minutes at 100% CPU</strong>. Based on code review, here's what likely happened:</p><table style=\"width:100%;border-collapse:collapse;margin-top:12px;\"><tr><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">What We Know</th><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Evidence</th></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">Process was Claude CLI</td><td style=\"padding:8px;color:#98c379;\">Command shown as 'claude' in top</td></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">Running 7:33 minutes</td><td style=\"padding:8px;color:#e5c07b;\">TIME+ column in screenshot</td></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">100% single-core CPU</td><td style=\"padding:8px;color:#f97316;\">101.7% CPU usage</td></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">Not Johnathon's session</td><td style=\"padding:8px;color:#61afef;\">Session ses_9de05b41-3ee completed at 23:20:49 (24s runtime)</td></tr></table></div><div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\"><h3 style=\"color:#c678dd;margin:0 0 12px 0;\">Most Likely Cause: Tool Execution Loop</h3><p style=\"margin:0 0 8px 0;\">Claude CLI was likely stuck in a <strong>tool execution loop</strong> - repeatedly calling tools that return partial results, triggering more calls.</p><p style=\"margin:0 0 8px 0;\">Examples that can cause this:</p><ul style=\"margin:0;padding-left:20px;line-height:1.8;color:#9ca3af;\"><li>Grep/Glob returning massive results → Claude tries to read each file</li><li>Task tool spawning subagents that spawn more subagents</li><li>Web searches that keep finding \"more to explore\"</li><li>Retry loops on failing tool calls</li></ul></div><div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\"><h3 style=\"color:#61afef;margin:0 0 12px 0;\">Current Safeguards (Already in Place)</h3><table style=\"width:100%;border-collapse:collapse;\"><tr><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Safeguard</th><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Setting</th><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Status</th></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">promptTimeoutMs</td><td style=\"padding:8px;color:#98c379;\">600,000ms (10 min)</td><td style=\"padding:8px;color:#98c379;\">✓ Working</td></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">Heartbeat events</td><td style=\"padding:8px;color:#98c379;\">Every 5 seconds</td><td style=\"padding:8px;color:#98c379;\">✓ Working</td></tr><tr><td style=\"padding:8px;color:#e0e0e0;\">Timeout resets on activity</td><td style=\"padding:8px;color:#e5c07b;\">resetTimeout() in stream.ts</td><td style=\"padding:8px;color:#f97316;\">⚠ PROBLEM</td></tr></table></div><div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;border-left:4px solid #f97316;\"><h3 style=\"color:#f97316;margin:0 0 12px 0;\">The Bug: Activity Resets Defeat Timeout</h3><p style=\"margin:0 0 12px 0;\">In <code style=\"background:#1e1e1e;padding:2px 6px;border-radius:4px;\">stream.ts:356</code>, <strong>every event from Claude resets the 10-minute timeout</strong>:</p><pre style=\"background:#1e1e1e;padding:12px;border-radius:4px;overflow-x:auto;font-size:0.9rem;margin:0;\"><span style=\"color:#7f848e;\">// Reset timeout on any activity from Claude</span>\n<span style=\"color:#e5c07b;\">onActivity</span>();</pre><p style=\"margin:12px 0 0 0;\">If Claude is in a loop calling tools, it's constantly emitting events → timeout <strong>never triggers</strong> → process runs forever.</p></div><div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\"><h3 style=\"color:#98c379;margin:0 0 12px 0;\">Proposed Fixes</h3><table style=\"width:100%;border-collapse:collapse;\"><tr><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Fix</th><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Description</th><th style=\"text-align:left;padding:8px;border-bottom:1px solid #444;color:#9ca3af;\">Priority</th></tr><tr><td style=\"padding:8px;color:#e5c07b;\">1. Hard max timeout</td><td style=\"padding:8px;\">Add <code style=\"background:#1e1e1e;padding:2px 4px;border-radius:3px;\">maxTurnTimeMs</code> (e.g., 15 min) that <strong>cannot be reset</strong></td><td style=\"padding:8px;color:#f97316;\">Critical</td></tr><tr><td style=\"padding:8px;color:#e5c07b;\">2. Tool call limit</td><td style=\"padding:8px;\">Add <code style=\"background:#1e1e1e;padding:2px 4px;border-radius:3px;\">maxToolCallsPerTurn</code> (e.g., 100) - stop after N tool calls</td><td style=\"padding:8px;color:#f97316;\">Critical</td></tr><tr><td style=\"padding:8px;color:#e5c07b;\">3. CPU monitoring</td><td style=\"padding:8px;\">Track process CPU time via <code style=\"background:#1e1e1e;padding:2px 4px;border-radius:3px;\">/proc/[pid]/stat</code>, kill if >5min CPU</td><td style=\"padding:8px;color:#e5c07b;\">High</td></tr><tr><td style=\"padding:8px;color:#e5c07b;\">4. Stall detection</td><td style=\"padding:8px;\">Detect <strong>same tool</strong> called >10x consecutively → likely stuck</td><td style=\"padding:8px;color:#61afef;\">Medium</td></tr></table></div><div style=\"background:#2d2d2d;padding:12px;border-radius:6px;margin-top:16px;\"><h4 style=\"margin:0 0 8px 0;color:#fff;\">Implementation Priority</h4><p style=\"color:#9ca3af;margin:0;\">Fixes #1 and #2 should be implemented immediately - they're simple config additions that prevent runaway processes.</p></div></div>",
    "requestedAt": "2026-01-05T09:25:00.000Z",
    "requestId": "bfa2a770-64f1-473b-aeea-d83f02f965bb",
    "choices": [
      {
        "label": "Implement fixes",
        "value": "Implement the hard max timeout and tool call limit fixes for konsole",
        "primary": true
      },
      {
        "label": "Add to backlog",
        "value": "Add these konsole stability fixes to the VIBE.md backlog for later implementation"
      },
      {
        "label": "More details",
        "value": "Show me more details about the current timeout code and how it could be improved"
      }
    ],
    "turnTiming": {
      "totalMs": 81633,
      "ttfeMs": 3794,
      "thinkingMs": 72550,
      "toolExecutionMs": 5286,
      "toolCallCount": 11,
      "thinkingPct": 89,
      "toolsPct": 6,
      "ttfePct": 5
    }
  },
  "createdBy": "claude",
  "createdAt": "2026-01-04T23:26:32.337Z",
  "updatedAt": "2026-01-04T23:26:41.224Z",
  "requestId": "bfa2a770-64f1-473b-aeea-d83f02f965bb",
  "scope": "vibetools",
  "tags": [
    "konsole",
    "performance",
    "stability",
    "architecture"
  ],
  "targetUser": "claude"
}